Healthcare analytics, AI solutions for biological big data, providing an AI platform for the biotech, life sciences, medical and pharmaceutical industries, as well as for related technological approaches, i.e., curation and text analysis with machine learning and other activities related to AI applications to these industries.
Building AI Is Hard—So Facebook Is Building AI That Builds AI
By forcing computers to do more of the grunt work, the world’s biggest tech companies are accelerating how quickly AI enters the everyday world.
In other words, for computers to get smarter faster, computers themselves must handle even more of the grunt work. The giants of the Internet are building computing systems that can test countless machine learning algorithms on behalf of their engineers, that can cycle through so many possibilities on their own. Better yet, these companies are building AI algorithms that can help build AI algorithms. No joke. Inside Facebook, engineers have designed what they like to call an “automated machine learning engineer,” an artificially intelligent system that helps create artificially intelligent systems. It’s a long way from perfection. But the goal is to create new AI models using as little human grunt work as possible.
Feeling the Flow
After Facebook’s $104 billion IPO in 2012, Hussein Mehanna and other engineers on the Facebook ads team felt an added pressure to improve the company’s ad targeting, to more precisely match ads to the hundreds of millions of people using its social network. This meant building deep neural networks and other machine learning algorithms that could make better use of the vast amounts of data Facebook collects on the characteristics and behavior of those hundreds of millions of people.
Artificial Intelligence is starting to turn invisible from the outside in — and vice versa. The exact effects and workings of AI technologies are becoming..
Following this, we are able to perceive manifestations and presentations of artificial intelligence, but the intelligence itself becomes unknowable to humans through human senses. Currently there are two distinct traits in this development.
First, most algorithmic systems, as well as the latest advancements in AI technologies, are black boxes; inaccessible, unfathomable and uncontrollable to most people.
Therefore, it’s hard to perceive or assess how intelligent systems shape your life online and offline, from your latest song recommendations to your personalized insurance policy, not to mention the algorithmic stock market trading that shapes the global market economy affecting almost every aspect of modern life.
Concretely, when the actions of intelligent systems become more holistically intertwined with personal, social, cultural, political and economical systems, it becomes challenging to distinguish the exact effects or impact of the machine intelligence itself.
Second, AI technologies are becoming so complex that they are hard to understand — even for the experts designing and developing them. In his recent book, The Master Algorithm, machine learning expert Pedro Domingos points out that already back in 1950s scientists created an algorithm that could do something that humans couldn’t fully comprehend.
This development hasn’t changed its course; rather, to the contrary. With the current pace of AI development, even seasoned experts have a hard time keeping up.
Today’s various machine learning systems can already provide unexpected insights in varying fields, from personalization technologies to particle physics, from cooking recipes and outlandish game moves to crime prevention and bioengineering. Concretely, specialized systems can empower scientific discoveries in biology or help you choose the best route to your next meeting.
Gene Editing with CRISPR gets Crisper, Volume 2 (Volume Two: Latest in Genomics Methodologies for Therapeutics: Gene Editing, NGS and BioInformatics, Simulations and the Genome Ontology), Part 2: CRISPR for Gene Editing and DNA Repair
Gene Editing with CRISPR gets Crisper
Curators: Larry H. Bernstein, MD, FCAP and Aviva Lev-Ari, PhD, RN
CRISPR Moves from Butchery to Surgery
More Genomes Are Going Under the CRISPR Knife, So Surgical Standards Are Rising
The Dharmacon subsidary of GE Healthcare provides the Edit-R Lentiviral Gene Engineering platform. It is based on the natural S. pyrogenes system, but unlike that system, which uses a single guide RNA (sgRNA), the platform uses two component RNAs, a gene-specific CRISPR RNA (crRNA) and a universal trans-activating crRNA (tracrRNA). Once hybridized to the universal tracrRNA (blue), the crRNA (green) directs the Cas9 nuclease to a specific genomic region to induce a double- strand break.
Scientists recently convened at the CRISPR Precision Gene Editing Congress, held in Boston, to discuss the new technology. As with any new technique, scientists have discovered that CRISPR comes with its own set of challenges, and the Congress focused its discussion around improving specificity, efficiency, and delivery.
In the naturally occurring system, CRISPR-Cas9 works like a self-vaccination in the bacterial immune system by targeting and cleaving viral DNA sequences stored from previous encounters with invading phages. The endogenous system uses two RNA elements, CRISPR RNA (crRNA) and trans-activating RNA (tracrRNA), which come together and guide the Cas9 nuclease to the target DNA.
Early publications that demonstrated CRISPR gene editing in mammalian cells combined the crRNA and tracrRNA sequences to form one long transcript called asingle-guide RNA (sgRNA). However, an alternative approach is being explored by scientists at the Dharmacon subsidiary of GE Healthcare. These scientists have a system that mimics the endogenous system through a synthetic two-component approach thatpreserves individual crRNA and tracrRNA. The tracrRNA is universal to any gene target or species; the crRNA contains the information needed to target the gene of interest.
Predesigned Guide RNAs
In contrast to sgRNAs, which are generated through either in vitro transcription of a DNA template or a plasmid-based expression system, synthetic crRNA and tracrRNA eliminate the need for additional cloning and purification steps. The efficacy of guide RNA (gRNA), whether delivered as a sgRNA or individual crRNA and tracrRNA, depends not only on DNA binding, but also on the generation of an indel that will deliver the coup de grâce to gene function.
“Almost all of the gRNAs were able to create a break in genomic DNA,” said Louise Baskin, senior product manager at Dharmacon. “But there was a very wide range in efficiency and in creating functional protein knock-outs.”
To remove the guesswork from gRNA design, Dharmacon developed an algorithm to predict gene knockout efficiency using wet-lab data. They also incorporated specificity as a component of their algorithm, using a much more comprehensive alignment tool to predict potential off-target effects caused by mismatches and bulges often missed by other alignment tools. Customers can enter their target gene to access predesigned gRNAs as either two-component RNAs or lentiviral sgRNA vectors for multiple applications.
“We put time and effort into our algorithm to ensure that our guide RNAs are not only functional but also highly specific,” asserts Baskin. “As a result, customers don’t have to do any design work.”
MilliporeSigma’s CRISPR Epigenetic Activator is based on fusion of a nuclease-deficient Cas9 (dCas9) to the catalytic histone acetyltransferase (HAT) core domain of the human E1A-associated protein p300. This technology allows researchers to target specific DNA regions or gene sequences. Researchers can localize epigenetic changes to their target of interest and see the effects of those changes in gene expression.
Knockout experiments are a powerful tool for analyzing gene function. However, for researchers who want to introduce DNA into the genome, guide design, donor DNA selection, and Cas9 activity are paramount to successful DNA integration.MilliporeSigma offers two formats for donor DNA: double-stranded DNA (dsDNA) plasmids and single-stranded DNA (ssDNA) oligonucleotides. The most appropriate format depends on cell type and length of the donor DNA. “There are some cell types that have immune responses to dsDNA,” said Gregory Davis, Ph.D., R&D manager, MilliporeSigma.
The ssDNA format can save researchers time and money, but it has a limited carrying capacity of approximately 120 base pairs.In addition to selecting an appropriate donor DNA format, controlling where, how, and when the Cas9 enzyme cuts can affect gene-editing efficiency. Scientists are playing tug-of-war, trying to pull cells toward the preferred homology-directed repair (HDR) and away from the less favored nonhomologous end joining (NHEJ) repair mechanism.One method to achieve this modifies the Cas9 enzyme to generate a nickase that cuts only one DNA strand instead of creating a double-strand break. Accordingly, MilliporeSigma has created a Cas9 paired-nickase system that promotes HDR, while also limiting off-target effects and increasing the number of sequences available for site-dependent gene modifications, such as disease-associated single nucleotide polymorphisms (SNPs).“The best thing you can do is to cut as close to the SNP as possible,” advised Dr. Davis. “As you move the double-stranded break away from the site of mutation you get an exponential drop in the frequency of recombination.”
Ribonucleo-protein Complexes
Another strategy to improve gene-editing efficiency, developed by Thermo Fisher, involves combining purified Cas9 protein with gRNA to generate a stable ribonucleoprotein (RNP) complex. In contrast to plasmid- or mRNA-based formats, which require transcription and/or translation, the Cas9 RNP complex cuts DNA immediately after entering the cell. Rapid clearance of the complex from the cell helps to minimize off-target effects, and, unlike a viral vector, the transient complex does not introduce foreign DNA sequences into the genome.
To deliver their Cas9 RNP complex to cells, Thermo Fisher has developed a lipofectamine transfection reagent called CRISPRMAX. “We went back to the drawing board with our delivery, screened a bunch of components, and got a brand-new, fully optimized lipid nanoparticle formulation,” explained Jon Chesnut, Ph.D., the company’s senior director of synthetic biology R&D. “The formulation is specifically designed for delivering the RNP to cells more efficiently.”
Besides the reagent and the formulation, Thermo Fisher has also developed a range of gene-editing tools. For example, it has introduced the Neon® transfection system for delivering DNA, RNA, or protein into cells via electroporation. Dr. Chesnut emphasized the company’s focus on simplifying complex workflows by optimizing protocols and pairing everything with the appropriate up- and downstream reagents.
From Mammalian Cells to Microbes
One of the first sources of CRISPR technology was the Feng Zhang laboratory at the Broad Institute, which counted among its first licensees a company called GenScript. This company offers a gene-editing service called GenCRISPR™ to establish mammalian cell lines with CRISPR-derived gene knockouts.
“There are a lot of challenges with mammalian cells, and each cell line has its own set of issues,” said Laura Geuss, a marketing specialist at GenScript. “We try to offer a variety of packages that can help customers who have difficult-to-work-with cells.” These packages include both viral-based and transient transfection techniques.
However, the most distinctive service offered by GenScript is its microbial genome-editing service for bacteria (Escherichia coli) and yeast (Saccharomyces cerevisiae). The company’s strategy for gene editing in bacteria can enable seamless knockins, knockouts, or gene replacements by combining CRISPR with lambda red recombineering. Traditionally one of the most effective methods for gene editing in microbes, recombineering allows editing without restriction enzymes through in vivo homologous recombination mediated by a phage-based recombination system such as lambda red.
On its own, lambda red technology cannot target multiple genes, but when paired with CRISPR, it allows the editing of multiple genes with greater efficiency than is possible with CRISPR alone, as the lambda red proteins help repair double-strand breaks in E. coli. The ability to knockout different gene combinations makes Genscript’s microbial editing service particularly well suited for the optimization of metabolic pathways.
Pooled and Arrayed Library Strategies
Scientists are using CRISPR technology for applications such as metabolic engineering and drug development. Yet another application area benefitting from CRISPR technology is cancer research. Here, the use of pooled CRISPR libraries is becoming commonplace. Pooled CRISPR libraries can help detect mutations that affect drug resistance, and they can aid in patient stratification and clinical trial design.
Pooled screening uses proliferation or viability as a phenotype to assess how genetic alterations, resulting from the application of a pooled CRISPR library, affect cell growth and death in the presence of a therapeutic compound. The enrichment or depletion of different gRNA populations is quantified using deep sequencing to identify the genomic edits that result in changes to cell viability.
MilliporeSigma provides pooled CRISPR libraries ranging from the whole human genome to smaller custom pools for these gene-function experiments. For pharmaceutical and biotech companies, Horizon Discovery offers a pooled screening service, ResponderSCREEN, which provides a whole-genome pooled screen to identify genes that confer sensitivity or resistance to a compound. This service is comprehensive, taking clients from experimental design all the way through to suggestions for follow-up studies.
Horizon Discovery maintains a Research Biotech business unit that is focused on target discovery and enabling translational medicine in oncology. “Our internal backbone gives us the ability to provide expert advice demonstrated by results,” said Jon Moore, Ph.D., the company’s CSO.
In contrast to a pooled screen, where thousands of gRNA are combined in one tube, an arrayed screen applies one gRNA per well, removing the need for deep sequencing and broadening the options for different endpoint assays. To establish and distribute a whole-genome arrayed lentiviral CRISPR library, MilliporeSigma partnered with the Wellcome Trust Sanger Institute. “This is the first and only arrayed CRISPR library in the world,” declared Shawn Shafer, Ph.D., functional genomics market segment manager, MilliporeSigma. “We were really proud to partner with Sanger on this.”
Pooled and arrayed screens are powerful tools for studying gene function. The appropriate platform for an experiment, however, will be determined by the desired endpoint assay.
The QX200 Droplet Digital PCR System from Bio-Rad Laboratories can provide researchers with an absolute measure of target DNA molecules for EvaGreen or probe-based digital PCR applications. The system, which can provide rapid, low-cost, ultra-sensitive quantification of both NHEJ- and HDR-editing events, consists of two instruments, the QX200 Droplet Generator and the QX200 Droplet Reader, and their associated consumables.
Finally, one last challenge for CRISPR lies in the detection and quantification of changes made to the genome post-editing. Conventional methods for detecting these alterations include gel methods and next-generation sequencing. While gel methods lack sensitivity and scalability, next-generation sequencing is costly and requires intensive bioinformatics.
To address this gap, Bio-Rad Laboratories developed a set of assay strategies to enable sensitive and precise edit detection with its Droplet Digital PCR (ddPCR) technology. The platform is designed to enable absolute quantification of nucleic acids with high sensitivity, high precision, and short turnaround time through massive droplet partitioning of samples.
Using a validated assay, a typical ddPCR experiment takes about five to six hours to complete. The ddPCR platform enables detection of rare mutations, and publications have reported detection of precise edits at a frequency of <0.05%, and of NHEJ-derived indels at a frequency as low as 0.1%. In addition to quantifying precise edits, indels, and computationally predicted off-target mutations, ddPCR can also be used to characterize the consequences of edits at the RNA level.
According to a recently published Science paper, the laboratory of Charles A. Gersbach, Ph.D., at Duke University used ddPCR in a study of muscle function in a mouse model of Duchenne muscular dystrophy. Specifically, ddPCR was used to assess the efficiency of CRISPR-Cas9 in removing the mutated exon 23 from the dystrophin gene. (Exon 23 deletion by CRISPR-Cas9 resulted in expression of the modified dystrophin gene and significant enhancement of muscle force.)
Quantitative ddPCR showed that exon 23 was deleted in ~2% of all alleles from the whole-muscle lysate. Further ddPCR studies found that 59% of mRNA transcripts reflected the deletion.
“There’s an overarching idea that the genome-editing field is moving extremely quickly, and for good reason,” asserted Jennifer Berman, Ph.D., staff scientist, Bio-Rad Laboratories. “There’s a lot of exciting work to be done, but detection and quantification of edits can be a bottleneck for researchers.”
The gene-editing field is moving quickly, and new innovations are finding their way into the laboratory as researchers lay the foundation for precise, well-controlled gene editing with CRISPR.
Researchers utilized a systems biology approach to develop new methods to assess drug sensitivity in cells. [The Institute for Systems Biology]
Understanding how cells respond and proliferate in the presence of anticancer compounds has been the foundation of drug discovery ideology for decades. Now, a new study from scientists at Vanderbilt University casts significant suspicion on the primary method used to test compounds for anticancer activity in cells—instilling doubt on methods employed by the entire scientific enterprise and pharmaceutical industry to discover new cancer drugs.
“More than 90% of candidate cancer drugs fail in late-stage clinical trials, costing hundreds of millions of dollars,” explained co-senior author Vito Quaranta, M.D., director of the Quantitative Systems Biology Center at Vanderbilt. “The flawed in vitro drug discovery metric may not be the only responsible factor, but it may be worth pursuing an estimate of its impact.”
The Vanderbilt investigators have developed what they believe to be a new metric for evaluating a compound’s effect on cell proliferation—called the DIP (drug-induced proliferation) rate—that overcomes the flawed bias in the traditional method.
The findings from this study were published recently in Nature Methods in an article entitled “An Unbiased Metric of Antiproliferative Drug Effect In Vitro.”
For more than three decades, researchers have evaluated the ability of a compound to kill cells by adding the compound in vitro and counting how many cells are alive after 72 hours. Yet, proliferation assays that measure cell number at a single time point don’t take into account the bias introduced by exponential cell proliferation, even in the presence of the drug.
“Cells are not uniform, they all proliferate exponentially, but at different rates,” Dr. Quaranta noted. “At 72 hours, some cells will have doubled three times and others will not have doubled at all.”
Dr. Quaranta added that drugs don’t all behave the same way on every cell line—for example, a drug might have an immediate effect on one cell line and a delayed effect on another.
The research team decided to take a systems biology approach, a mixture of experimentation and mathematical modeling, to demonstrate the time-dependent bias in static proliferation assays and to develop the time-independent DIP rate metric.
“Systems biology is what really makes the difference here,” Dr. Quaranta remarked. “It’s about understanding cells—and life—as dynamic systems.”This new study is of particular importance in light of recent international efforts to generate data sets that include the responses of thousands of cell lines to hundreds of compounds. Using the
Cancer Cell Line Encyclopedia (CCLE) and
Genomics of Drug Sensitivity in Cancer (GDSC) databases
will allow drug discovery scientists to include drug response data along with genomic and proteomic data that detail each cell line’s molecular makeup.
“The idea is to look for statistical correlations—these particular cell lines with this particular makeup are sensitive to these types of compounds—to use these large databases as discovery tools for new therapeutic targets in cancer,” Dr. Quaranta stated. “If the metric by which you’ve evaluated the drug sensitivity of the cells is wrong, your statistical correlations are basically no good.”
The Vanderbilt team evaluated the responses from four different melanoma cell lines to the drug vemurafenib, currently used to treat melanoma, with the standard metric—used for the CCLE and GDSC databases—and with the DIP rate. In one cell line, they found a glaring disagreement between the two metrics.
“The static metric says that the cell line is very sensitive to vemurafenib. However, our analysis shows this is not the case,” said co-lead study author Leonard Harris, Ph.D., a systems biology postdoctoral fellow at Vanderbilt. “A brief period of drug sensitivity, quickly followed by rebound, fools the static metric, but not the DIP rate.”
Dr. Quaranta added that the findings “suggest we should expect melanoma tumors treated with this drug to come back, and that’s what has happened, puzzling investigators. DIP rate analyses may help solve this conundrum, leading to better treatment strategies.”
The researchers noted that using the DIP rate is possible because of advances in automation, robotics, microscopy, and image processing. Moreover, the DIP rate metric offers another advantage—it can reveal which drugs are truly cytotoxic (cell killing), rather than merely cytostatic (cell growth inhibiting). Although cytostatic drugs may initially have promising therapeutic effects, they may leave tumor cells alive that then have the potential to cause the cancer to recur.
The Vanderbilt team is currently in the process of identifying commercial entities that can further refine the software and make it widely available to the research community to inform drug discovery.
An unbiased metric of antiproliferative drug effect in vitro
In vitro cell proliferation assays are widely used in pharmacology, molecular biology, and drug discovery. Using theoretical modeling and experimentation, we show that current metrics of antiproliferative small molecule effect suffer from time-dependent bias, leading to inaccurate assessments of parameters such as drug potency and efficacy. We propose the drug-induced proliferation (DIP) rate, the slope of the line on a plot of cell population doublings versus time, as an alternative, time-independent metric.
Researchers develop a technique to direct chromosome recombination with CRISPR/Cas9, allowing high-resolution genetic mapping of phenotypic traits in yeast.
Researchers used CRISPR/Cas9 to make a targeted double-strand break (DSB) in one arm of a yeast chromosome labeled with a green fluorescent protein (GFP) gene. A within-cell mechanism called homologous repair (HR) mends the broken arm using its homolog, resulting in a recombined region from the site of the break to the chromosome tip. When this cell divides by mitosis, each daughter cell will contain a homozygous section in an outcome known as “loss of heterozygosity” (LOH). One of the daughter cells is detectable because, due to complete loss of the GFP gene, it will no longer be fluorescent.REPRINTED WITH PERMISSION FROM M.J. SADHU ET AL., SCIENCE
When mapping phenotypic traits to specific loci, scientists typically rely on the natural recombination of chromosomes during meiotic cell division in order to infer the positions of responsible genes. But recombination events vary with species and chromosome region, giving researchers little control over which areas of the genome are shuffled. Now, a team at the University of California, Los Angeles (UCLA), has found a way around these problems by using CRISPR/Cas9 to direct targeted recombination events during mitotic cell division in yeast. The team described its technique today (May 5) in Science.
“Current methods rely on events that happen naturally during meiosis,” explained study coauthor Leonid Kruglyak of UCLA. “Whatever rate those events occur at, you’re kind of stuck with. Our idea was that using CRISPR, we can generate those events at will, exactly where we want them, in large numbers, and in a way that’s easy for us to pull out the cells in which they happened.”
Generally, researchers use coinheritance of a trait of interest with specific genetic markers—whose positions are known—to figure out what part of the genome is responsible for a given phenotype. But the procedure often requires impractically large numbers of progeny or generations to observe the few cases in which coinheritance happens to be disrupted informatively. What’s more, the resolution of mapping is limited by the length of the smallest sequence shuffled by recombination—and that sequence could include several genes or gene variants.
“Once you get down to that minimal region, you’re done,” said Kruglyak. “You need to switch to other methods to test every gene and every variant in that region, and that can be anywhere from challenging to impossible.”
But programmable, DNA-cutting champion CRISPR/Cas9 offered an alternative. During mitotic—rather than meiotic—cell division, rare, double-strand breaks in one arm of a chromosome preparing to split are sometimes repaired by a mechanism called homologous recombination. This mechanism uses the other chromosome in the homologous pair to replace the sequence from the break down to the end of the broken arm. Normally, such mitotic recombination happens so rarely as to be impractical for mapping purposes. With CRISPR/Cas9, however, the researchers found that they could direct double-strand breaks to any locus along a chromosome of interest (provided it was heterozygous—to ensure that only one of the chromosomes would be cut), thus controlling the sites of recombination.
Combining this technique with a signal of recombination success, such as a green fluorescent protein (GFP) gene at the tip of one chromosome in the pair, allowed the researchers to pick out cells in which recombination had occurred: if the technique failed, both daughter cells produced by mitotic division would be heterozygous, with one copy of the signal gene each. But if it succeeded, one cell would end up with two copies, and the other cell with none—an outcome called loss of heterozygosity.
“If we get loss of heterozygosity . . . half the cells derived after that loss of heterozygosity event won’t have GFP anymore,” study coauthor Meru Sadhu of UCLA explained. “We search for these cells that don’t have GFP out of the general population of cells.” If these non-fluorescent cells with loss of heterozygosity have the same phenotype as the parent for a trait of interest, then CRISPR/Cas9-targeted recombination missed the responsible gene. If the phenotype is affected, however, then the trait must be linked to a locus in the recombined, now-homozygous region, somewhere between the cut site and the GFP gene.
By systematically making cuts using CRISPR/Cas9 along chromosomes in a hybrid, diploid strain ofSaccharomyces cerevisiae yeast, picking out non-fluorescent cells, and then observing the phenotype, the UCLA team demonstrated that it could rapidly identify the phenotypic contribution of specific gene variants. “We can simply walk along the chromosome and at every [variant] position we can ask, does it matter for the trait we’re studying?” explained Kruglyak.
For example, the team showed that manganese sensitivity—a well-defined phenotypic trait in lab yeast—could be pinpointed using this method to a single nucleotide polymorphism (SNP) in a gene encoding the Pmr1 protein (a manganese transporter).
Jason Moffat, a molecular geneticist at the University of Toronto who was not involved in the work, toldThe Scientist that researchers had “dreamed about” exploiting these sorts of mechanisms for mapping purposes, but without CRISPR, such techniques were previously out of reach. Until now, “it hasn’t been so easy to actually make double-stranded breaks on one copy of a pair of chromosomes, and then follow loss of heterozygosity in mitosis,” he said, adding that he hopes to see the approach translated into human cell lines.
Applying the technique beyond yeast will be important, agreed cell and developmental biologist Ethan Bier of the University of California, San Diego, because chromosomal repair varies among organisms. “In yeast, they absolutely demonstrate the power of [this method],” he said. “We’ll just have to see how the technology develops in other systems that are going to be far less suited to the technology than yeast. . . . I would like to see it implemented in another system to show that they can get the same oomph out of it in, say, mammalian somatic cells.”
Kruglyak told The Scientist that work in higher organisms, though planned, is still in early stages; currently, his team is working to apply the technique to map loci responsible for trait differences between—rather than within—yeast species.
“We have a much poorer understanding of the differences across species,” Sadhu explained. “Except for a few specific examples, we’re pretty much in the dark there.”
Linkage and association studies have mapped thousands of genomic regions that contribute to phenotypic variation, but narrowing these regions to the underlying causal genes and variants has proven much more challenging. Resolution of genetic mapping is limited by the recombination rate. We developed a method that uses CRISPR to build mapping panels with targeted recombination events. We tested the method by generating a panel with recombination events spaced along a yeast chromosome arm, mapping trait variation, and then targeting a high density of recombination events to the region of interest. Using this approach, we fine-mapped manganese sensitivity to a single polymorphism in the transporter Pmr1. Targeting recombination events to regions of interest allows us to rapidly and systematically identify causal variants underlying trait differences.
Thank you, David, for the kind words and comments. We agree that the most immediate applications of the CRISPR-based recombination mapping will be in unicellular organisms and cell culture. We also think the method holds a lot of promise for research in multicellular organisms, although we did not mean to imply that it “will be an efficient mapping method for all multicellular organisms”. Every organism will have its own set of constraints as well as experimental tools that will be relevant when adapting a new technique. To best help experts working on these organisms, here are our thoughts on your questions.
You asked about mutagenesis during recombination. We Sanger sequenced 72 of our LOH lines at the recombination site and did not observe any mutations, as described in the supplementary materials. We expect the absence of mutagenesis is because we targeted heterozygous sites where the untargeted allele did not have a usable PAM site; thus, following LOH, the targeted site is no longer present and cutting stops. In your experiments you targeted sites that were homozygous; thus, following recombination, the CRISPR target site persisted, and continued cutting ultimately led to repair by NHEJ and mutagenesis.
As to the more general question of the optimal mapping strategies in different organisms, they will depend on the ease of generating and screening for editing events, the cost and logistics of maintaining and typing many lines, and generation time, among other factors. It sounds like in Drosophila today, your related approach of generating markers with CRISPR, and then enriching for natural recombination events that separate them, is preferable. In yeast, we’ve found the opposite to be the case. As you note, even in Drosophila, our approach may be preferable for regions with low or highly non-uniform recombination rates.
Finally, mapping in sterile interspecies hybrids should be straightforward for unicellular hybrids (of which there are many examples) and for cells cultured from hybrid animals or plants. For studies in hybrid multicellular organisms, we agree that driving mitotic recombination in the early embryo may be the most promising approach. Chimeric individuals with mitotic clones will be sufficient for many traits. Depending on the system, it may in fact be possible to generate diploid individuals with uniform LOH genotype, but this is certainly beyond the scope of our paper. The calculation of the number of lines assumes that the mapping is done in a single step; as you note in your earlier comment, mapping sequentially can reduce this number dramatically.
This is a lovely method and should find wide applicability in many settings, especially for microorganisms and cell lines. However, it is not clear that this approach will be, as implied by the discussion, an efficient mapping method for all multicellular organisms. I have performed similar experiments in Drosophila, focused on meiotic recombination, on a much smaller scale, and found that CRISPR-Cas9 can indeed generate targeted recombination at gRNA target sites. In every case I tested, I found that the recombination event was associated with a deletion at the gRNA site, which is probably unimportant for most mapping efforts, but may be a concern in some specific cases, for example for clinical applications. It would be interesting to know how often mutations occurred at the targeted gRNA site in this study.
The wider issue, however, is whether CRISPR-mediated recombination will be more efficient than other methods of mapping. After careful consideration of all the costs and the time involved in each of the steps for Drosophila, we have decided that targeted meiotic recombination using flanking visible markers will be, in most cases, considerably more efficient than CRISPR-mediated recombination. This is mainly due to the large expense of injecting embryos and the extensive effort and time required to screen injected animals for appropriate events. It is both cheaper and faster to generate markers (with CRISPR) and then perform a large meiotic recombination mapping experiment than it would be to generate the lines required for CRISPR-mediated recombination mapping. It is possible to dramatically reduce costs by, for example, mapping sequentially at finer resolution. But this approach would require much more time than marker-assisted mapping. If someone develops a rapid and cheap method of reliably introducing DNA into Drosophila embryos, then this calculus might change.
However, it is possible to imagine situations where CRISPR-mediated mapping would be preferable, even for Drosophila. For example, some genomic regions display extremely low or highly non-uniform recombination rates. It is possible that CRISPR-mediated mapping could provide a reasonable approach to fine mapping genes in these regions.
The authors also propose the exciting possibility that CRISPR-mediated loss of heterozygosity could be used to map traits in sterile species hybrids. It is not entirely obvious to me how this experiment would proceed and I hope the authors can illuminate me. If we imagine driving a recombination event in the early embryo (with maternal Cas9 from one parent and gRNA from a second parent), then at best we would end up with chimeric individuals carrying mitotic clones. I don’t think one could generate diploid animals where all cells carried the same loss of heterozygosity event. Even if we could, this experiment would require construction of a substantial number of stable transgenic lines expressing gRNAs. Mapping an ~20Mbp chromosome arm to ~10kb would require on the order of two-thousand transgenic lines. Not an undertaking to be taken lightly. It is already possible to perform similar tests (hemizygosity tests) using D. melanogaster deficiency lines in crosses with D. simulans, so perhaps CRISPR-mediated LOH could complement these deficiency screens for fine mapping efforts. But, at the moment, it is not clear to me how to do the experiment.
Imaging of Cancer Cells, Volume 2 (Volume Two: Latest in Genomics Methodologies for Therapeutics: Gene Editing, NGS and BioInformatics, Simulations and the Genome Ontology), Part 1: Next Generation Sequencing (NGS)
Imaging of Cancer Cells
Larry H. Bernstein, MD, FCAP, Curator
LPBI
Microscope uses nanosecond-speed laser and deep learning to detect cancer cells more efficiently
April 13, 2016
Scientists at the California NanoSystems Institute at UCLA have developed a new technique for identifying cancer cells in blood samples faster and more accurately than the current standard methods.
In one common approach to testing for cancer, doctors add biochemicals to blood samples. Those biochemicals attach biological “labels” to the cancer cells, and those labels enable instruments to detect and identify them. However, the biochemicals can damage the cells and render the samples unusable for future analyses. There are other current techniques that don’t use labeling but can be inaccurate because they identify cancer cells based only on one physical characteristic.
Time-stretch quantitative phase imaging (TS-QPI) and analytics system
The new technique images cells without destroying them and can identify 16 physical characteristics — including size, granularity and biomass — instead of just one.
The new technique combines two components that were invented at UCLA:
A “photonic time stretch” microscope, which is capable of quickly imaging cells in blood samples. Invented by Barham Jalali, professor and Northrop-Grumman Optoelectronics Chair in electrical engineering, it works by taking pictures of flowing blood cells using laser bursts (similar to how a camera uses a flash). Each flash only lasts nanoseconds (billionths of a second) to avoid damage to cells, but that normally means the images are both too weak to be detected and too fast to be digitized by normal instrumentation. The new microscope overcomes those challenges by using specially designed optics that amplify and boost the clarity of the images, and simultaneously slow them down enough to be detected and digitized at a rate of 36 million images per second.
A deep learning computer program, which identifies cancer cells with more than 95 percent accuracy. Deep learning is a form of artificial intelligence that uses complex algorithms to extract patterns and knowledge from rich multidimenstional datasets, with the goal of achieving accurate decision making.
The study was published in the open-access journal Nature Scientific Reports. The researchers write in the paper that the system could lead to data-driven diagnoses by cells’ physical characteristics, which could allow quicker and earlier diagnoses of cancer, for example, and better understanding of the tumor-specific gene expression in cells, which could facilitate new treatments for disease.
The research was supported by NantWorks, LLC.
Abstract of Deep Learning in Label-free Cell Classification
Label-free cell analysis is essential to personalized genomics, cancer diagnostics, and drug development as it avoids adverse effects of staining reagents on cellular viability and cell signaling. However, currently available label-free cell assays mostly rely only on a single feature and lack sufficient differentiation. Also, the sample size analyzed by these assays is limited due to their low throughput. Here, we integrate feature extraction and deep learning with high-throughput quantitative imaging enabled by photonic time stretch, achieving record high accuracy in label-free cell classification. Our system captures quantitative optical phase and intensity images and extracts multiple biophysical features of individual cells. These biophysical measurements form a hyperdimensional feature space in which supervised learning is performed for cell classification. We compare various learning algorithms including artificial neural network, support vector machine, logistic regression, and a novel deep learning pipeline, which adopts global optimization of receiver operating characteristics. As a validation of the enhanced sensitivity and specificity of our system, we show classification of white blood T-cells against colon cancer cells, as well as lipid accumulating algal strains for biofuel production. This system opens up a new path to data-driven phenotypic diagnosis and better understanding of the heterogeneous gene expressions in cells.
references:
Claire Lifan Chen, Ata Mahjoubfar, Li-Chia Tai, Ian K. Blaby, Allen Huang, Kayvan Reza Niazi & Bahram Jalali. Deep Learning in Label-free Cell Classification. Scientific Reports 6, Article number: 21471 (2016); doi:10.1038/srep21471 (open access)
Supplementary Information
Deep Learning in Label-free Cell Classification
Claire Lifan Chen, Ata Mahjoubfar, Li-Chia Tai, Ian K. Blaby, Allen Huang,Kayvan Reza Niazi & Bahram Jalali
Deep learning extracts patterns and knowledge from rich multidimenstional datasets. While it is extensively used for image recognition and speech processing, its application to label-free classification of cells has not been exploited. Flow cytometry is a powerful tool for large-scale cell analysis due to its ability to measure anisotropic elastic light scattering of millions of individual cells as well as emission of fluorescent labels conjugated to cells1,2. However, each cell is represented with single values per detection channels (forward scatter, side scatter, and emission bands) and often requires labeling with specific biomarkers for acceptable classification accuracy1,3. Imaging flow cytometry4,5 on the other hand captures images of cells, revealing significantly more information about the cells. For example, it can distinguish clusters and debris that would otherwise result in false positive identification in a conventional flow cytometer based on light scattering6.
In addition to classification accuracy, the throughput is another critical specification of a flow cytometer. Indeed high throughput, typically 100,000 cells per second, is needed to screen a large enough cell population to find rare abnormal cells that are indicative of early stage diseases. However there is a fundamental trade-off between throughput and accuracy in any measurement system7,8. For example, imaging flow cytometers face a throughput limit imposed by the speed of the CCD or the CMOS cameras, a number that is approximately 2000 cells/s for present systems9. Higher flow rates lead to blurred cell images due to the finite camera shutter speed. Many applications of flow analyzers such as cancer diagnostics, drug discovery, biofuel development, and emulsion characterization require classification of large sample sizes with a high-degree of statistical accuracy10. This has fueled research into alternative optical diagnostic techniques for characterization of cells and particles in flow.
Recently, our group has developed a label-free imaging flow-cytometry technique based on coherent optical implementation of the photonic time stretch concept11. This instrument overcomes the trade-off between sensitivity and speed by using Amplified Time-stretch Dispersive Fourier Transform12,13,14,15. In time stretched imaging16, the object’s spatial information is encoded in the spectrum of laser pulses within a pulse duration of sub-nanoseconds (Fig. 1). Each pulse representing one frame of the camera is then stretched in time so that it can be digitized in real-time by an electronic analog-to-digital converter (ADC). The ultra-fast pulse illumination freezes the motion of high-speed cells or particles in flow to achieve blur-free imaging. Detection sensitivity is challenged by the low number of photons collected during the ultra-short shutter time (optical pulse width) and the drop in the peak optical power resulting from the time stretch. These issues are solved in time stretch imaging by implementing a low noise-figure Raman amplifier within the dispersive device that performs time stretching8,11,16. Moreover, warped stretch transform17,18can be used in time stretch imaging to achieve optical image compression and nonuniform spatial resolution over the field-of-view19. In the coherent version of the instrument, the time stretch imaging is combined with spectral interferometry to measure quantitative phase and intensity images in real-time and at high throughput20. Integrated with a microfluidic channel, coherent time stretch imaging system in this work measures both quantitative optical phase shift and loss of individual cells as a high-speed imaging flow cytometer, capturing 36 million images per second in flow rates as high as 10 meters per second, reaching up to 100,000 cells per second throughput.
Box 1: The pulse train is spatially dispersed into a train of rainbow flashes illuminating the target as line scans. The spatial features of the target are encoded into the spectrum of the broadband optical pulses, each representing a one-dimensional frame. The ultra-short optical pulse illumination freezes the motion of cells during high speed flow to achieve blur-free imaging with a throughput of 100,000 cells/s. The phase shift and intensity loss at each location within the field of view are embedded into the spectral interference patterns using a Michelson interferometer. Box 2: The interferogram pulses were then stretched in time so that spatial information could be mapped into time through time-stretch dispersive Fourier transform (TS-DFT), and then captured by a single pixel photodetector and an analog-to-digital converter (ADC). The loss of sensitivity at high shutter speed is compensated by stimulated Raman amplification during time stretch. Box 3: (a) Pulse synchronization; the time-domain signal carrying serially captured rainbow pulses is transformed into a series of one-dimensional spatial maps, which are used for forming line images. (b) The biomass density of a cell leads to a spatially varying optical phase shift. When a rainbow flash passes through the cells, the changes in refractive index at different locations will cause phase walk-off at interrogation wavelengths. Hilbert transformation and phase unwrapping are used to extract the spatial phase shift. (c) Decoding the phase shift in each pulse at each wavelength and remapping it into a pixel reveals the protein concentration distribution within cells. The optical loss induced by the cells, embedded in the pulse intensity variations, is obtained from the amplitude of the slowly varying envelope of the spectral interferograms. Thus, quantitative optical phase shift and intensity loss images are captured simultaneously. Both images are calibrated based on the regions where the cells are absent. Cell features describing morphology, granularity, biomass, etc are extracted from the images. (d) These biophysical features are used in a machine learning algorithm for high-accuracy label-free classification of the cells.
On another note, surface markers used to label cells, such as EpCAM21, are unavailable in some applications; for example, melanoma or pancreatic circulating tumor cells (CTCs) as well as some cancer stem cells are EpCAM-negative and will escape EpCAM-based detection platforms22. Furthermore, large-population cell sorting opens the doors to downstream operations, where the negative impacts of labels on cellular behavior and viability are often unacceptable23. Cell labels may cause activating/inhibitory signal transduction, altering the behavior of the desired cellular subtypes, potentially leading to errors in downstream analysis, such as DNA sequencing and subpopulation regrowth. In this way, quantitative phase imaging (QPI) methods24,25,26,27 that categorize unlabeled living cells with high accuracy are needed. Coherent time stretch imaging is a method that enables quantitative phase imaging at ultrahigh throughput for non-invasive label-free screening of large number of cells.
In this work, the information of quantitative optical loss and phase images are fused into expert designed features, leading to a record label-free classification accuracy when combined with deep learning. Image mining techniques are applied, for the first time, to time stretch quantitative phase imaging to measure biophysical attributes including protein concentration, optical loss, and morphological features of single cells at an ultrahigh flow rate and in a label-free fashion. These attributes differ widely28,29,30,31 among cells and their variations reflect important information of genotypes and physiological stimuli32. The multiplexed biophysical features thus lead to information-rich hyper-dimensional representation of the cells for label-free classification with high statistical precision.
We further improved the accuracy, repeatability, and the balance between sensitivity and specificity of our label-free cell classification by a novel machine learning pipeline, which harnesses the advantages of multivariate supervised learning, as well as unique training by evolutionary global optimization of receiver operating characteristics (ROC). To demonstrate sensitivity, specificity, and accuracy of multi-feature label-free flow cytometry using our technique, we classified (1) OT-IIhybridoma T-lymphocytes and SW-480 colon cancer epithelial cells, and (2) Chlamydomonas reinhardtii algal cells (herein referred to as Chlamydomonas) based on their lipid content, which is related to the yield in biofuel production. Our preliminary results show that compared to classification by individual biophysical parameters, our label-free hyperdimensional technique improves the detection accuracy from 77.8% to 95.5%, or in other words, reduces the classification inaccuracy by about five times. ……..
Feature Extraction
The decomposed components of sequential line scans form pairs of spatial maps, namely, optical phase and loss images as shown in Fig. 2 (see Section Methods: Image Reconstruction). These images are used to obtain biophysical fingerprints of the cells8,36. With domain expertise, raw images are fused and transformed into a suitable set of biophysical features, listed in Table 1, which the deep learning model further converts into learned features for improved classification.
The new technique combines two components that were invented at UCLA:
A “photonic time stretch” microscope, which is capable of quickly imaging cells in blood samples. Invented by Barham Jalali, professor and Northrop-Grumman Optoelectronics Chair in electrical engineering, it works by taking pictures of flowing blood cells using laser bursts (similar to how a camera uses a flash). Each flash only lasts nanoseconds (billionths of a second) to avoid damage to cells, but that normally means the images are both too weak to be detected and too fast to be digitized by normal instrumentation. The new microscope overcomes those challenges by using specially designed optics that amplify and boost the clarity of the images, and simultaneously slow them down enough to be detected and digitized at a rate of 36 million images per second.
A deep learning computer program, which identifies cancer cells with more than 95 percent accuracy. Deep learning is a form of artificial intelligence that uses complex algorithms to extract patterns and knowledge from rich multidimenstional datasets, with the goal of achieving accurate decision making.
The study was published in the open-access journal Nature Scientific Reports. The researchers write in the paper that the system could lead to data-driven diagnoses by cells’ physical characteristics, which could allow quicker and earlier diagnoses of cancer, for example, and better understanding of the tumor-specific gene expression in cells, which could facilitate new treatments for disease.
The optical loss images of the cells are affected by the attenuation of multiplexed wavelength components passing through the cells. The attenuation itself is governed by the absorption of the light in cells as well as the scattering from the surface of the cells and from the internal cell organelles. The optical loss image is derived from the low frequency component of the pulse interferograms. The optical phase image is extracted from the analytic form of the high frequency component of the pulse interferograms using Hilbert Transformation, followed by a phase unwrapping algorithm. Details of these derivations can be found in Section Methods. Also, supplementary Videos 1 and 2 show measurements of cell-induced optical path length difference by TS-QPI at four different points along the rainbow for OT-II and SW-480, respectively.
Table 1: List of extracted features.
Feature Name Description Category
Figure 3: Biophysical features formed by image fusion.
(a) Pairwise correlation matrix visualized as a heat map. The map depicts the correlation between all major 16 features extracted from the quantitative images. Diagonal elements of the matrix represent correlation of each parameter with itself, i.e. the autocorrelation. The subsets in box 1, box 2, and box 3 show high correlation because they are mainly related to morphological, optical phase, and optical loss feature categories, respectively. (b) Ranking of biophysical features based on their AUCs in single-feature classification. Blue bars show performance of the morphological parameters, which includes diameter along the interrogation rainbow, diameter along the flow direction, tight cell area, loose cell area, perimeter, circularity, major axis length, orientation, and median radius. As expected, morphology contains most information, but other biophysical features can contribute to improved performance of label-free cell classification. Orange bars show optical phase shift features i.e. optical path length differences and refractive index difference. Green bars show optical loss features representing scattering and absorption by the cell. The best performed feature in these three categories are marked in red.
Figure 4: Machine learning pipeline. Information of quantitative optical phase and loss images are fused to extract multivariate biophysical features of each cell, which are fed into a fully-connected neural network.
The neural network maps input features by a chain of weighted sum and nonlinear activation functions into learned feature space, convenient for classification. This deep neural network is globally trained via area under the curve (AUC) of the receiver operating characteristics (ROC). Each ROC curve corresponds to a set of weights for connections to an output node, generated by scanning the weight of the bias node. The training process maximizes AUC, pushing the ROC curve toward the upper left corner, which means improved sensitivity and specificity in classification.
…. How to cite this article: Chen, C. L. et al. Deep Learning in Label-free Cell Classification.
To better characterize the functional context of genomic variations in cancer, researchers developed a new computer algorithm called REVEALER. [UC San Diego Health]
Scientists at the University of California San Diego School of Medicine and the Broad Institute say they have developed a new computer algorithm—REVEALER—to better characterize the functional context of genomic variations in cancer. The tool, described in a paper (“Characterizing Genomic Alterations in Cancer by Complementary Functional Associations”) published in Nature Biotechnology, is designed to help researchers identify groups of genetic variations that together associate with a particular way cancer cells get activated, or how they respond to certain treatments.
REVEALER is available for free to the global scientific community via the bioinformatics software portal GenePattern.org.
“This computational analysis method effectively uncovers the functional context of genomic alterations, such as gene mutations, amplifications, or deletions, that drive tumor formation,” said senior author Pablo Tamayo, Ph.D., professor and co-director of the UC San Diego Moores Cancer Center Genomics and Computational Biology Shared Resource.
Dr. Tamayo and team tested REVEALER using The Cancer Genome Atlas (TCGA), the NIH’s database of genomic information from more than 500 human tumors representing many cancer types. REVEALER revealed gene alterations associated with the activation of several cellular processes known to play a role in tumor development and response to certain drugs. Some of these gene mutations were already known, but others were new.
For example, the researchers discovered new activating genomic abnormalities for beta-catenin, a cancer-promoting protein, and for the oxidative stress response that some cancers hijack to increase their viability.
REVEALER requires as input high-quality genomic data and a significant number of cancer samples, which can be a challenge, according to Dr. Tamayo. But REVEALER is more sensitive at detecting similarities between different types of genomic features and less dependent on simplifying statistical assumptions, compared to other methods, he adds.
“This study demonstrates the potential of combining functional profiling of cells with the characterizations of cancer genomes via next-generation sequencing,” said co-senior author Jill P. Mesirov, Ph.D., professor and associate vice chancellor for computational health sciences at UC San Diego School of Medicine.
Characterizing genomic alterations in cancer by complementary functional associations
Jong Wook Kim, Olga B Botvinnik, Omar Abudayyeh, Chet Birger, et al.
Systematic efforts to sequence the cancer genome have identified large numbers of mutations and copy number alterations in human cancers. However, elucidating the functional consequences of these variants, and their interactions to drive or maintain oncogenic states, remains a challenge in cancer research. We developed REVEALER, a computational method that identifies combinations of mutually exclusive genomic alterations correlated with functional phenotypes, such as the activation or gene dependency of oncogenic pathways or sensitivity to a drug treatment. We used REVEALER to uncover complementary genomic alterations associated with the transcriptional activation of β-catenin and NRF2, MEK-inhibitor sensitivity, and KRAS dependency. REVEALER successfully identified both known and new associations, demonstrating the power of combining functional profiles with extensive characterization of genomic alterations in cancer genomes
Figure 2: REVEALER results for transcriptional activation of β-catenin in cancer.close
(a) This heatmap illustrates the use of the REVEALER approach to find complementary genomic alterations that match the transcriptional activation of β-catenin in cancer. The target profile is a TCF4 reporter that provides an estimate of…
An imaging-based platform for high-content, quantitative evaluation of therapeutic response in 3D tumour models
Jonathan P. Celli, Imran Rizvi, Adam R. Blanden, Iqbal Massodi, Michael D. Glidden, Brian W. Pogue & Tayyaba Hasan
While it is increasingly recognized that three-dimensional (3D) cell culture models recapitulate drug responses of human cancers with more fidelity than monolayer cultures, a lack of quantitative analysis methods limit their implementation for reliable and routine assessment of emerging therapies. Here, we introduce an approach based on computational analysis of fluorescence image data to provide high-content readouts of dose-dependent cytotoxicity, growth inhibition, treatment-induced architectural changes and size-dependent response in 3D tumour models. We demonstrate this approach in adherent 3D ovarian and pancreatic multiwell extracellular matrix tumour overlays subjected to a panel of clinically relevant cytotoxic modalities and appropriately designed controls for reliable quantification of fluorescence signal. This streamlined methodology reads out the high density of information embedded in 3D culture systems, while maintaining a level of speed and efficiency traditionally achieved with global colorimetric reporters in order to facilitate broader implementation of 3D tumour models in therapeutic screening.
The attrition rates for preclinical development of oncology therapeutics are particularly dismal due to a complex set of factors which includes 1) the failure of pre-clinical models to recapitulate determinants of in vivo treatment response, and 2) the limited ability of available assays to extract treatment-specific data integral to the complexities of therapeutic responses1,2,3. Three-dimensional (3D) tumour models have been shown to restore crucial stromal interactions which are missing in the more commonly used 2D cell culture and that influence tumour organization and architecture4,5,6,7,8, as well as therapeutic response9,10, multicellular resistance (MCR)11,12, drug penetration13,14, hypoxia15,16, and anti-apoptotic signaling17. However, such sophisticated models can only have an impact on therapeutic guidance if they are accompanied by robust quantitative assays, not only for cell viability but also for providing mechanistic insights related to the outcomes. While numerous assays for drug discovery exist18, they are generally not developed for use in 3D systems and are often inherently unsuitable. For example, colorimetric conversion products have been noted to bind to extracellular matrix (ECM)19 and traditional colorimetric cytotoxicity assays reduce treatment response to a single number reflecting a biochemical event that has been equated to cell viability (e.g. tetrazolium salt conversion20). Such approaches fail to provide insight into the spatial patterns of response within colonies, morphological or structural effects of drug response, or how overall culture viability may be obscuring the status of sub-populations that are resistant or partially responsive. Hence, the full benefit of implementing 3D tumour models in therapeutic development has yet to be realized for lack of analytical methods that describe the very aspects of treatment outcome that these systems restore.
Motivated by these factors, we introduce a new platform for quantitative in situ treatment assessment (qVISTA) in 3D tumour models based on computational analysis of information-dense biological image datasets (bioimage-informatics)21,22. This methodology provides software end-users with multiple levels of complexity in output content, from rapidly-interpreted dose response relationships to higher content quantitative insights into treatment-dependent architectural changes, spatial patterns of cytotoxicity within fields of multicellular structures, and statistical analysis of nodule-by-nodule size-dependent viability. The approach introduced here is cognizant of tradeoffs between optical resolution, data sampling (statistics), depth of field, and widespread usability (instrumentation requirement). Specifically, it is optimized for interpretation of fluorescent signals for disease-specific 3D tumour micronodules that are sufficiently small that thousands can be imaged simultaneously with little or no optical bias from widefield integration of signal along the optical axis of each object. At the core of our methodology is the premise that the copious numerical readouts gleaned from segmentation and interpretation of fluorescence signals in these image datasets can be converted into usable information to classify treatment effects comprehensively, without sacrificing the throughput of traditional screening approaches. It is hoped that this comprehensive treatment-assessment methodology will have significant impact in facilitating more sophisticated implementation of 3D cell culture models in preclinical screening by providing a level of content and biological relevance impossible with existing assays in monolayer cell culture in order to focus therapeutic targets and strategies before costly and tedious testing in animal models.
Using two different cell lines and as depicted in Figure 1, we adopt an ECM overlay method pioneered originally for 3D breast cancer models23, and developed in previous studies by us to model micrometastatic ovarian cancer19,24. This system leads to the formation of adherent multicellular 3D acini in approximately the same focal plane atop a laminin-rich ECM bed, implemented here in glass-bottom multiwell imaging plates for automated microscopy. The 3D nodules resultant from restoration of ECM signaling5,8, are heterogeneous in size24, in contrast to other 3D spheroid methods, such as rotary or hanging drop cultures10, in which cells are driven to aggregate into uniformly sized spheroids due to lack of an appropriate substrate to adhere to. Although the latter processes are also biologically relevant, it is the adherent tumour populations characteristic of advanced metastatic disease that are more likely to be managed with medical oncology, which are the focus of therapeutic evaluation herein. The heterogeneity in 3D structures formed via ECM overlay is validated here by endoscopic imaging ofin vivo tumours in orthotopic xenografts derived from the same cells (OVCAR-5).
Figure 1: A simplified schematic flow chart of imaging-based quantitative in situ treatment assessment (qVISTA) in 3D cell culture.
(This figure was prepared in Adobe Illustrator® software by MD Glidden, JP Celli and I Rizvi). A detailed breakdown of the image processing (Step 4) is provided in Supplemental Figure 1.
A critical component of the imaging-based strategy introduced here is the rational tradeoff of image-acquisition parameters for field of view, depth of field and optical resolution, and the development of image processing routines for appropriate removal of background, scaling of fluorescence signals from more than one channel and reliable segmentation of nodules. In order to obtain depth-resolved 3D structures for each nodule at sub-micron lateral resolution using a laser-scanning confocal system, it would require ~ 40 hours (at approximately 100 fields for each well with a 20× objective, times 1 minute/field for a coarse z-stack, times 24 wells) to image a single plate with the same coverage achieved in this study. Even if the resources were available to devote to such time-intensive image acquisition, not to mention the processing, the optical properties of the fluorophores would change during the required time frame for image acquisition, even with environmental controls to maintain culture viability during such extended imaging. The approach developed here, with a mind toward adaptation into high throughput screening, provides a rational balance of speed, requiring less than 30 minutes/plate, and statistical rigour, providing images of thousands of nodules in this time, as required for the high-content analysis developed in this study. These parameters can be further optimized for specific scenarios. For example, we obtain the same number of images in a 96 well plate as for a 24 well plate by acquiring only a single field from each well, rather than 4 stitched fields. This quadruples the number conditions assayed in a single run, at the expense of the number of nodules per condition, and therefore the ability to obtain statistical data sets for size-dependent response, Dfrac and other segmentation-dependent numerical readouts.
We envision that the system for high-content interrogation of therapeutic response in 3D cell culture could have widespread impact in multiple arenas from basic research to large scale drug development campaigns. As such, the treatment assessment methodology presented here does not require extraordinary optical instrumentation or computational resources, making it widely accessible to any research laboratory with an inverted fluorescence microscope and modestly equipped personal computer. And although we have focused here on cancer models, the methodology is broadly applicable to quantitative evaluation of other tissue models in regenerative medicine and tissue engineering. While this analysis toolbox could have impact in facilitating the implementation of in vitro 3D models in preclinical treatment evaluation in smaller academic laboratories, it could also be adopted as part of the screening pipeline in large pharma settings. With the implementation of appropriate temperature controls to handle basement membranes in current robotic liquid handling systems, our analyses could be used in ultra high-throughput screening. In addition to removing non-efficacious potential candidate drugs earlier in the pipeline, this approach could also yield the additional economic advantage of minimizing the use of costly time-intensive animal models through better estimates of dose range, sequence and schedule for combination regimens.
Microscope Uses AI to Find Cancer Cells More Efficiently
Scientists at the California NanoSystems Institute at UCLA have developed a new technique for identifying cancer cells in blood samples faster and more accurately than the current standard methods.
In one common approach to testing for cancer, doctors add biochemicals to blood samples. Those biochemicals attach biological “labels” to the cancer cells, and those labels enable instruments to detect and identify them. However, the biochemicals can damage the cells and render the samples unusable for future analyses.
There are other current techniques that don’t use labeling but can be inaccurate because they identify cancer cells based only on one physical characteristic.
The new technique images cells without destroying them and can identify 16 physical characteristics — including size, granularity and biomass — instead of just one. It combines two components that were invented at UCLA: a photonic time stretch microscope, which is capable of quickly imaging cells in blood samples, and a deep learning computer program that identifies cancer cells with over 95 percent accuracy.
Deep learning is a form of artificial intelligence that uses complex algorithms to extract meaning from data with the goal of achieving accurate decision making.
The study, which was published in the journal Nature Scientific Reports, was led by Barham Jalali, professor and Northrop-Grumman Optoelectronics Chair in electrical engineering; Claire Lifan Chen, a UCLA doctoral student; and Ata Mahjoubfar, a UCLA postdoctoral fellow.
Photonic time stretch was invented by Jalali, and he holds a patent for the technology. The new microscope is just one of many possible applications; it works by taking pictures of flowing blood cells using laser bursts in the way that a camera uses a flash. This process happens so quickly — in nanoseconds, or billionths of a second — that the images would be too weak to be detected and too fast to be digitized by normal instrumentation.
The new microscope overcomes those challenges using specially designed optics that boost the clarity of the images and simultaneously slow them enough to be detected and digitized at a rate of 36 million images per second. It then uses deep learning to distinguish cancer cells from healthy white blood cells.
“Each frame is slowed down in time and optically amplified so it can be digitized,” Mahjoubfar said. “This lets us perform fast cell imaging that the artificial intelligence component can distinguish.”
Normally, taking pictures in such minuscule periods of time would require intense illumination, which could destroy live cells. The UCLA approach also eliminates that problem.
“The photonic time stretch technique allows us to identify rogue cells in a short time with low-level illumination,” Chen said.
The researchers write in the paper that the system could lead to data-driven diagnoses by cells’ physical characteristics, which could allow quicker and earlier diagnoses of cancer, for example, and better understanding of the tumor-specific gene expression in cells, which could facilitate new treatments for disease. ….. see also http://www.nature.com/article-assets/npg/srep/2016/160315/srep21471/images_hires/m685/srep21471-f1.jpg
Can machines be creative? Recent successes in AI have shown that machines can now perform at human levels in many tasks that, just a few years ago, were considered to be decades away, like driving cars, understanding spoken language, and recognizing objects. But these are all tasks where we know what needs to be done, and the machine is just imitating us. What about tasks where the right answers are not known? Can machines be programmed to find solutions on their own, and perhaps even come up with creative solutions that humans would find difficult?
The answer is a definite yes! There are branches of AI focused precisely on this challenge, including evolutionary computation and reinforcement learning. Like the popular deep learning methods, which are responsible for many of the recent AI successes, these branches of AI have benefitted from the million-fold increase in computing power we’ve seen over the last two decades. There arenow antennas in spacecraft so complex they could only be designed through computational evolution. There are game playing agents in Othello, Backgammon, and most recently in Go that have learned to play at the level of the best humans, and in the case of AlphaGo, even beyond the ability of the best humans. There are non-player characters in Unreal Tournament that have evolved to be indistinguishable from humans, thereby passing the Turing test— at least for game bots. And in finance, there are computational traders in the stock market evolved to make real money.
Many new applications have suddenly come within our reach thanks to computational creativity — even though most of us do not realize it yet. If you are facing a design problem where potential solutions can be tested automatically, chances are you could evolve those solutions automatically as well. In areas where computers are already used to draft designs, the natural next step is to harness evolutionary search. This will allow human designers to gain more traction for their ideas, such as machine parts that are easier to manufacture, stock portfolios that minimize risk, or websites that result in more conversions. In other areas, it may take some engineering effort to define the design problem for the computer, but the effort may be rewarded by truly novel designs, such as finless rockets, new video game genres, personalized preventive medicine, and safer and more efficient traffic.
How do we address medical diagnostic errors? Volume 2 (Volume Two: Latest in Genomics Methodologies for Therapeutics: Gene Editing, NGS and BioInformatics, Simulations and the Genome Ontology), Part 1: Next Generation Sequencing (NGS)
How do we address medical diagnostic errors?
Larry H. Bernstein, MD, FCAP, Curator
LPBI
It is my observation over the course of 50 years in medicine that there is no simple resolution to a frequently identified problem, which has grown more serious with the implementation of hospital-wide and system-wide health information technology. There has been reduction of hospital staff, nurses and physicians, with an increased patient load even while the technology has advanced significantly. Over this investment in technology and manpower the hospital staff time has become dependent on the medical record and a large portion of their time is drawn away from patient care. This is not a substantial improvement in the last decade. The problem has been a lack of focus on the correct design of these systems. They should be interactive and they should have incorporated anticipatory logic. Instead, they are not that much more than large electronic filing cabinets. There are large sources of information in laboratory, radiology and pharmacy and are siloed. When one drives a car, one does not need to look under the hood, a sentiment that has been expressed by a real expert.
March 25, 2016 DARK DAILYinfo@darkreport.comvia cmail2.com
Along with its assessment of the rate of errors in diagnosis, the IOM has a plan to improve, but will doctors accept the IOM’s advice, or continue business as usual?
Diagnostic errors in the American healthcare system is a problem that is now on the radar screen of policymakers at the Institute of Medicine (IOM). Pathologists andclinical laboratory professionals will welcome this development, because recommendations from the IOM carry weight with Congress.
Thus, should the IOM develop specific actions items intended to reduce medical errors, not only are these suggestions likely to involve more effective use of medical laboratory tests by physicians, but there is a strong probability that Congress might eventually write these recommendations into future healthcare legislation.
The Institute of Medicine is a division of the National Academies of Sciences, Engineering, and Medicine. The IOM recently convened a committee that released a list of recommendations to address the problem of diagnostic errors in medicine. Those recommendations, however, are running up against ingrained mindsets and overconfidence on the part of physicians who are reluctant to include decision-support technology in the diagnostic process.
Misdiagnoses in healthcare has led to tragic consequences. Over the years, the IOM has worked to increase the public’s awareness of the problem of misdiagnosis. It has also encouraged health information technology (HIT) developers to provide clinicians with the decision-support systems they need to prevent misdiagnosis from happening in the first place. But getting physicians to use the new tools has become a challenge.
IOM Dedicated to Improving Diagnostic Accuracy
Preventing diagnostic errors has been a primary target in medical reform for many years. A 1999 report by the U.S Institute of Medicine (IOM) titled: “To Err Is Human: Building a Safer Health System” first drew attention to the rate of preventable medical errors in the U.S.
Since then, much has been written about the problem of diagnostic inaccuracy. Although many provider organizations are working to improve it, progress has been slow.
Edward Hoffer, MD (above), FACP, FACC, FACMI, is a faculty member at the Massachusetts General Hospital Laboratory of Computer Science. He told Modern Healthcare that getting doctors to use diagnostic software is a hard sell. “The main problem we face is trying to convince physicians that they actually need to use it,” he concluded. (Image copyright: Massachusetts General Hospital, Laboratory of Computer Science.)
Solutions for Preventing Errors Face Many Challenges
From the first evaluation of the patient, to the clinical laboratory, to physician follow-up; errors can occur throughout the chain of care. And a culture that discourages reporting of medical errors can make it very difficult for healthcare organizations to resolve these issues.
For example, an investigation by the Milwaukee Journal Sentinel uncovered what they say is a “secretive system [that] hides [medical] lab errors from the public and puts patients at risk.”
To make matters worse, according to the Journal Sentinel article, federal agencies that are tasked with oversight of the more than 35,000 clinical laboratories in the U.S. (such as the Joint Commission) are overloaded and thus may be missing many violations that would lead to sanctions or outright loss of accreditation.
The Milwaukee Journal Sentinel stated that, “even when serious violations are identified, offending labs are rarely sanctioned except in the most extreme cases.” And that “in 2013, just 90 sanctions were issued—accounting for not even 1% of the 35,000 labs that do high-level lab testing in the United States.”
Additionally, there are different types of misdiagnoses, which also complicates matters. A study by three professors at the Dartmouth Center for Healthcare Deliver Science investigated the cost of what they termed “silent misdiagnoses.”
According to a Dartmouth Now article outlining the study, the disparity between “the treatments patients want and what doctors think they want” accounts for a great number of misdiagnoses. The study’s authors call these misdiagnoses “silent” because they are rarely reported or recorded.
Albert G. Mulley, Jr., MD, MPP (above), is Managing Director, Global Health Care Delivery Science, The Dartmouth Institute for Health Policy and Clinical Practice; Professor of Medicine, Geisel School of Medicine at Dartmouth; and the lead author of the Dartmouth study. Mulley says that listening to the patient is more important than ever. “Today, the rise in treatment options makes this even more critical, not only to reach a correct medical diagnosis but also to understand fully the patients’ preferences, and reduce the huge waste in time and money that comes from delivery of services that patients often neither want nor need.” (Photo copyright: Dartmouth Now.)
IOM Committee on Diagnostic Error in Healthcare Releases Recommendations
This newest report is titled, “Improving Diagnosis in HealthCare.” It comes from the IOM’s National Academies of Medicine’s Committee on Diagnostic Error in Healthcare. The study lists recommendations that the IOM says will help improve the rate of diagnostic errors. Included are:
• Facilitate more effective teamwork in the diagnostic process;
• Enhance education in the healthcare community on the topic of improving diagnostics;
• Incorporate diagnostic support into electronic health records (EHRs), and make EHRs fully interoperable;
• Set up systems to identify, learn from, and reduce diagnostic errors;
• Establish a culture that encourages investigation of mistakes;
• Develop a collaborative reporting system that involves clinicians, government agencies, and liability insurers that allows everyone to learn from diagnostic errors and near misses;
• Design a payment system that supports correct diagnosis; and
• Provide funding for research into how to improve the diagnostic process.
One area of development that offers real hope for lowering the rate of misdiagnosis is medical decision-support software—specifically in clinical informatics, where the integration of health data with health information technology (HIT) supports physician decision making at the point of care.
Whether that software is part of an organization’s EHR system, or is a standalone program dedicated to reducing diagnostic errors, can make a difference. For example, a research study published in theJournal of Clinical Bioinformatics) suggests that health information technology has an important role to play in improving diagnostic accuracy.
The challenge is that, although the technology currently exists, diagnostic error rates remain high. A research study published in the Journal of the American Medical Association Internal Medicine (JAMA) in 2013 may explain why. It states that overconfidence could prevent some physicians from re-evaluating questionable diagnoses.
In a Modern Healthcare article, Edward Hoffer, MD, FACP, FACC, FACMI, a faculty member at the Massachusetts General Hospital Laboratory of Computer Science, said that getting doctors to use diagnostic software is a hard sell. “The main problem we face is trying to convince physicians that they actually need to use it,” he concluded.
All of the building blocks for improving the diagnostic process are either in place, or as in the case of EHR interoperability, in the works. Clinical laboratory personnel are uniquely positioned to assist physicians in improving through communication and the careful use of technology.
All of the building blocks for improving the diagnostic process are either in place, or as in the case of EHR interoperability, in the works. Nothing could be further from reality than what has been stated. There is no minimization of keystrokes. There is no anticipatory logic. There is no automated comparison of medication and diagnoses electronically. There is no check to determine the consistency of the laboratory results with a probability of diagnosis, or automated identification of tests that would clarify the situation.
Microsoft is carrying out a massive social experiment in China — and almost no one outside the country knows about it
Reporter: Aviva Lev-Ari, PhD, RN
Microsoft’s Xiaoice, which means “little Bing” in Chinese, has been described as the “largest Turing test in history” by a top researcher at the company.
TomorrowWorld: A world where everyone has a robot, or why 2040 could blow your mind
Reporter: Aviva Lev-Ari, PhD, RN
Technological change is accelerating today at an unprecedented speed and could create a world we can barely begin to imagine.
In March 2001, futurist Ray Kurzweil published an essay arguing that humans found it hard to comprehend their own future. It was clear from history, he argued, that technological change is exponential — even though most of us are unable to see it — and that in a few decades, the world would be unrecognizably different. “We won’t experience 100 years of progress in the 21st century — it will be more like 20,000 years of progress (at today’s rate),” he wrote, in ‘The Law of Accelerating Returns’.
Fifteen years on, Kurzweil is a director of engineering at Google and his essay has acquired a cult following among futurists. Some of its predictions are outlandish or over-hyped — but technology experts say that its basic tenets often hold. The evidence, they say, lies in the exponential advances in a suite of enabling technologies ranging from computing power to data storage, to the scale and performance of the Internet (see ‘Onwards and upwards’). These advances are creating tipping points — moments at which technologies such as robotics, artificial intelligence (AI), biology, nanotechnology and 3D printing cross a threshold and trigger sudden and significant change. “We live in a mind-blowingly different world than our grandparents,” says Fei-Fei Li, head of the Stanford Artificial Intelligence Laboratory in California, and this will be all the more true for our children and grandchildren (see ‘Future focus’).
Kurzweil and others have argued that people find this pace of change almost impossible to grasp, because it is human nature to perceive rates of progress as linear, not exponential — much as when one zooms in on a small part of a circle and it appears as an almost straight line. People tend to focus on the past few years, but pulling back reveals a much more dramatic change. Many things that society now takes for granted would have seemed like futuristic nonsense just a few decades ago. We can search across billions of pages, images and videos on the web; mobile phones have become ubiquitous; billions of connected smart sensors monitor in real time everything from the state of the planet to our heartbeats, sleep and steps; and drones and satellites the size of shoeboxes roam the skies.
If the pace of change is exponentially speeding up, all those advances could begin to look trivial within a few years. Take ‘deep learning’, a form of artificial intelligence that uses powerful microprocessor chips and algorithms to simulate neural networks that train and learn through experience, using massive data sets. Last month, the Google-owned AI company DeepMind used deep learning to enable a computer to beat for the first time a human professional at the game of Go, long considered one of the grand challenges of AI. Researchers told Nature that they foresee a future just 20 years from now — or even sooner — in which robots with AI are as common as cars or phones and are integrated into families, offices and factories. The “disruptive exponentials” of technological change will create “a world where everybody can have a robot and robots are pervasively integrated in the fabric of life”, says Daniela Rus, head of the Computer Science and Artificial Intelligence Laboratory at the Massachusetts Institute of Technology in Cambridge.
After decades in development, applications of AI are moving into the real world, says Li, with the arrival of self-driving cars, virtual reality and more. Progress in AI and robotics is likely to accelerate rapidly as deep-pocketed companies such as Google, Apple, Facebook and Microsoft pour billions of dollars into these fields. Gill Pratt, former head of the US Defense Advanced Research Projects Agency’s Robotics Challenge, asked last year whether robotics is about to undergo a ‘Cambrian explosion’ — a period of rapid machine diversification (G. A. PrattJ. Econ. Perspect. 29, 51–60; 2015). Although a single robot cannot yet match the learning ability of a toddler, Pratt pointed out that robots have one huge advantage: humans can communicate with each other at only 10 bits per second — whereas robots can communicate through the Internet at speeds 100 million times faster. This could, he said, result in multitides of robots building on each other’s learning experiences at lightning speed. Pratt was hired last September to head the Toyota Research Institute, a new US$1-billion AI and robotics research venture headquartered in Palo Alto, California.
Many researchers say that it is important to prepare for this new world. “We need to become much more responsible in terms of designing and operating these robots as they become more powerful,” says Li. In January 2015, a group including Elon Musk, Bill Gates and Stephen Hawking penned an open letter calling for extensive research to maximize the benefits of AI and avoid its potential pitfalls. The letter has now been signed by more than 8,000 people.
The superhero of artificial intelligence: can this genius keep it in check?
Reporter: Aviva Lev-Ari, PhD, RN
With his company DeepMind, Londoner Demis Hassabis is leading Google’s project to build software more powerful than the human brain. But what will this mean for the future of humankind?
My first encounter with Hassabis was back in the summer of 2014, a few months after the DeepMind acquisition. Since then, I’ve observed him at work in a variety of environments and have interviewed him formally for this profile on three separate occasions over the past eight months. In that time I’ve watched him evolve from Google’s AI genius to a compelling communicator who has found an effective way to describe to non-scientists like me his vastly complex work – about which he is infectiously passionate – and why it matters. Unpretentious and increasingly personable, he is very good at breaking down DeepMind’s approach; namely their combining of old and new AI techniques – such as, in Go, using traditional “tree search” methods for analysing moves with modern “deep neural networks”, which approximate the web of neurons in the brain – and also their methodical “marriage” of different areas of AI research.
2012pharmaceutical
sjwilliamspa
This is a lovely method and should find wide applicability in many settings, especially for microorganisms and cell lines. However, it is not clear that this approach will be, as implied by the discussion, an efficient mapping method for all multicellular organisms. I have performed similar experiments in Drosophila, focused on meiotic recombination, on a much smaller scale, and found that CRISPR-Cas9 can indeed generate targeted recombination at gRNA target sites. In every case I tested, I found that the recombination event was associated with a deletion at the gRNA site, which is probably unimportant for most mapping efforts, but may be a concern in some specific cases, for example for clinical applications. It would be interesting to know how often mutations occurred at the targeted gRNA site in this study.
The wider issue, however, is whether CRISPR-mediated recombination will be more efficient than other methods of mapping. After careful consideration of all the costs and the time involved in each of the steps for Drosophila, we have decided that targeted meiotic recombination using flanking visible markers will be, in most cases, considerably more efficient than CRISPR-mediated recombination. This is mainly due to the large expense of injecting embryos and the extensive effort and time required to screen injected animals for appropriate events. It is both cheaper and faster to generate markers (with CRISPR) and then perform a large meiotic recombination mapping experiment than it would be to generate the lines required for CRISPR-mediated recombination mapping. It is possible to dramatically reduce costs by, for example, mapping sequentially at finer resolution. But this approach would require much more time than marker-assisted mapping. If someone develops a rapid and cheap method of reliably introducing DNA into Drosophila embryos, then this calculus might change.
However, it is possible to imagine situations where CRISPR-mediated mapping would be preferable, even for Drosophila. For example, some genomic regions display extremely low or highly non-uniform recombination rates. It is possible that CRISPR-mediated mapping could provide a reasonable approach to fine mapping genes in these regions.
The authors also propose the exciting possibility that CRISPR-mediated loss of heterozygosity could be used to map traits in sterile species hybrids. It is not entirely obvious to me how this experiment would proceed and I hope the authors can illuminate me. If we imagine driving a recombination event in the early embryo (with maternal Cas9 from one parent and gRNA from a second parent), then at best we would end up with chimeric individuals carrying mitotic clones. I don’t think one could generate diploid animals where all cells carried the same loss of heterozygosity event. Even if we could, this experiment would require construction of a substantial number of stable transgenic lines expressing gRNAs. Mapping an ~20Mbp chromosome arm to ~10kb would require on the order of two-thousand transgenic lines. Not an undertaking to be taken lightly. It is already possible to perform similar tests (hemizygosity tests) using D. melanogaster deficiency lines in crosses with D. simulans, so perhaps CRISPR-mediated LOH could complement these deficiency screens for fine mapping efforts. But, at the moment, it is not clear to me how to do the experiment.