Posts Tagged ‘microarray’

Issues in Personalized Medicine: Discussions of Intratumor Heterogeneity from the Oncology Pharma forum on LinkedIn

Curator and Writer: Stephen J. Williams, Ph.D.

In an earlier post entitled “Issues in Personalized Medicine in Cancer: Intratumor Heterogeneity and Branched Evolution Revealed by Multiregion Sequencing” the heterogenic nature of solid tumors was discussed.  There resulted an excellent discussion in the Oncology Pharma forum on LinkedIn so I curated the comments (below article) to foster further discussion. To summarize the original post, this was a discussion of Dr. Charles Swanton’s paper[1] in which he and colleagues had noticed that individual biopsies from primary renal tumors displayed a variety of mutations of the same and different tumor suppressor genes (TSG), thereby not only revealing the heterogeneity of individual tumors but also how tumors can evolve.  Thus it was suggested that individual cells of a primary tumor can represent individual clones, each evolving on a distinct pathway to tumorigenicity and metastasis as each clone would have accumulated different passenger mutations.  It is these passenger mutations which have been posited to be responsible for a tumor’s continued growth (as discussed in the following post Rewriting the Mathematics of Tumor Growth; Teams Use Math Models to Sort Drivers from Passengers).  Indeed, as Dr. Swanton mentioned in the posting that it is very likely a solid tumor contains discrete clones with different driver and passenger mutations and possibly different mutated TSG but also this intra-tumor heterogeneity would have great implications for personalized chemotherapeutic strategies, not only against the primary tumor but against resistant outgrowth clones, and to the metastatic disease, as Swanton and colleagues had found that the metastatic disease displayed tremendously increased genomic instability than the underlying primary disease.

Therefore it may behoove the clinical oncologist to view solid tumors as a collection of multiple clones, each having their own mutagenic spectrum and tumorigenic phenotype.  Each of these clones may acquire further mutations which provide growth advantage over other clones in the early primary tumor.  In addition, branched evolution of a clone most likely depends more on genomic instability and epigenetic factors than on solely somatic mutation.

This is echoed in a  report in Carcinogenesis back in 2005[3] Lorena Losi, Benedicte Baisse, Hanifa Bouzourene and Jean Benhatter had shown some similar results in colorectal cancer as their abstract described:

“In primary colorectal cancers (CRCs), intratumoral genetic heterogeneity was more often observed in early than in advanced stages, at 90 and 67%, respectively. All but one of the advanced CRCs were composed of one predominant clone and other minor clones, whereas no predominant clone has been identified in half of the early cancers. A reduction of the intratumoral genetic heterogeneity for point mutations and a relative stability of the heterogeneity for allelic losses indicate that, during the progression of CRC, clonal selection and chromosome instability continue, while an increase cannot be proven.”

Therefore if a tumor had evolved in time closer to the initial driver mutation multiple therapies may be warranted while tumors which had not yet evolved much from their driver mutation may be tackled with an agent directed against that driver, hence the branched evolution as shown in the following diagram:

branced chain evolution cancer

Cancer Sequencing

Unravels clonal evolution.

From Carlos Caldas. (2012).

Nature Biotechnology V.30

pp 405-410.[2] used with









An article written by Drs. Andrei Krivtsov and Scott Armstron entitled “Can One Cell Influence Cancer Heterogeneity”[4] commented on a study by Friedman-Morvinski[5] in Inder Verma’s laboratory discussed how genetic lesions can revert differentiated neorons and glial cells to an undifferentiated state [an important phenotype in development of glioblastoma multiforme].

In particular it is discussed that epigenetic state of the transformed cell may contribute to the heterogeneity of the resultant tumor.  Indeed many investigators (initially discovered and proposed by Dr. Beatrice Mintz of the Institute for Cancer Research, later to be named the Fox Chase Cancer Center) show the cellular microenvironment influences transformation and tumor development[6-8].

Briefly the Friedman-Morvinski study used intra-cerebral ventricular (ICV) injection of lentivirus to introduce oncogenes within the CNS and produced tumors of multiple cell origins including neuronal and glial cell origin (neuroblastoma and glioma).  The important takeaway was differentiated somatic cells which acquire genetic lesions can transform to form multiple tumor types.  As the authors state, “cellular differentiation and specialization are accompanied by gradual changes in epigenetic programs” and that “the cell of origin may influence the epigenetic state of the tumor”.   In essence this means that the success of therapy may depend on the cellular state (whether stem cell, progenitor cell, or differentiated specialized cell) at time of transformation.  In other words tumors arising from cells with an epigenetic state seen in stem cells would be more resistant to therapy unless given an epigenetic therapy, such as azacytididne, retinoic acid or HDAC inhibitors.


So as the Oncology Pharma forum on LinkedIn was such an excellent discussion I would like to post the comments for curation purposes and foster further discussion.  I would like to thank everyone’s great comments below.  I would especially like to thank Dr. Emanuel Petricoin from George Mason and Dr. David Anderson for supplying extra papers which will be the subject of a future post. I had curated each comment with inserted LIVE LINKS to make it easier to refer to a paper and/or company mentioned in the comment.

The comments seemed to center on three main themes:

  1. 1.      Clinicians pondering the benefit to mutational spectrum analysis to determine personalized therapy and develop biomarkers of early disease
  2. 2.      A shift in the clinicians paradigm of cancer development, diagnoses, and treatment from strictly histologic evaluation to a genetic and altered cellular pathway view
  3. 3.      Use of proteomics, microarray and epigenetics as an alternative to mutational analysis to determine aberrant cellular networks in various stages of tumor development


Victor Levenson • Thanks for posting this! To be honest, I am puzzled by the insistence on sequencing as a tool for tumor analysis – we know that expression patterns rather than mutations in a limited number of genes determine tumor physiology (or, even more, physiology of any tissue). Since the AACR-2012 we know that different tumors have similar or even identical mutations, so >functional< rather than >structural< differences are important. Frankly, I’d be much more excited learning about expression pattern heterogeneity in tumors.Granted that is much more challenging than NGS sequencing, but the value of the data would be incomparable, especially in its application to biomarker development.

Stephen J. Williams, Ph.D. • Dear Dr. Levenson, thanks for your comments. I agree with you and in no way am insisting on the releiance of sequencing mutations in cancer as the sole means for determining therapy. It is extremely true that tumors will show tremendous heterogeneity of mRNA expression. There are a number of studies (one which I will post on that individual tumor cells will have differing expression patterns based on the levels of regional hypoxia within the tumor as well as other microenvironmental factors. I do have two posts on on this matter, curating various programs around the world which are using microarray expression analysis of tumors to determine personalized strategies. I believe the reliance on mutational analysis is based on the drugs that have been developed (such as Gleevec and crizotinib) which are based on mutant forms of BCR-Abl and ALK, respectively. However (as per two posts I did based on Mike Martin on our site “Mathematical Models of Driver and Passenger mutations) where he discusses how certain driver mutations will get the senescent cell over the hump to get to fully transformed and contribute to a certain level of growth while subsequent passengers are responsible for the sustained survival and expansion of the tumor.

Victor Levenson • Dr. Williams, thanks for the comments. Driving a senescent cell into proliferative stage is a tremendous change, which >may< begin with a mutation, but involves dramatic restructuring of transcription patterns that will drive the process. Hypoxia will definitely contribute to variations in the patterns, although will probably not be the main driver of the process. As to whether a mutation or a change in transcription pattern initiate the process, I am not sure we will ever be able to determine <grin>.

Vanisree Staniforth • Thanks for posting! Certainly a thought provoking article with regard to the future of personalized cancer therapies.


Dr. Raj Batra • If we follow Dr Levenson’s proposed conceptual approach (which we also published in 2009 and 2010), we are MUCH more likely to significantly impact tumor morbidity and mortality.

Stephen J. Williams, Ph.D. • Thanks Vanisiree and Dr. Batra for your comments. Hopefully we will see, from the future cancer statistics, how personlized therapy have improved outcomes for the solid tumors, like the hematologic cancers. 26 days ago

Emanuel Petricoin • The issue about intra and inter tumor heterogeneity is very important however since it is unknown which mutations are true drivers, an explanation of the results found in these studies simply could be the variances are all in the inconsequential mutations and the commonality is the driver mutations. Moreover, at the end of the day, its not the mRNA expression that we really care about but the functional protein signaling -phosphoprotein driven signaling architecture, that we care about since these are the drug targets directly.

Mohammad Azhar Aziz,PhD • This article addresses the potential complexity of dealing with cancer which is apparently increasing proportionally with the amount of data generated. Intratumor heterogeneity will remain there and even multiple biopsies that are randomly chosen will offer no conclusive solution.Mutations,expression profiles and functional protein signaling (as discussed above) alone can not provide any breakthrough. It will be a composite picture of all these and many other components (e.g. microenvironment, alternative splicing, epigenetics,non-coding RNAs etc.) that will hold the promises in the future. We have made phenomenal advances in understanding each of these aspects separately but definitely lack the tools to integrate all these. Developing tools to integrate all these data may provide some breakthrough in understanding and thus treating cancer.

Emanuel Petricoin • I agree Mohammad in a systems biology approach however the current compendium of drugs largely are kinase inhibitors or enzymatic inhibitors. Since most studies have shown little correlation between gene mutation and protein levels and phosphoprotein levels, for example, it is no wonder why the recent spate of failed trials (e.g. stratification by PIK3CA mutation or PTEN mutation for AKT-mTOR inhibitors) should come as any shock. We will be publishing work using protein pathway activation mapping coupled to laser dissection of a number of intra and inter tumoral analysis that indicates that the signaling architecture appears much more stable.

Stephen J. Williams, Ph.D. • Thank you Dr. Pettricoin for your comments. I eagerly await the publication of your results concerning proteomic evaluation of multiple biopsies of a tumor. I am very interested that you found limited intratuoral heterogeneity of signaling pathways given the diversity of intratumoral microenvironmental stresses (changes in regional hypoxia, blood flow, and populations of cancer stem cells). I agree with you and Mohammed that proteomic profiling will be imperative in determining personalized approaches for targeted therapy. Dr. Swanton had informed me that they had used IHC to determine if mTOR signaling had correlated with the mutational spectrum they had seen. In addition he had mentioned that there was enhanced genomic instability in the metastatic disease relative to the primary tumor and it would be very interesting to see how signaling pathways change in cohorts of matched metastatic and primary tumors. A few years ago we were looking at genes which were completely lost upon transformation of ovarian epithelial cells and worked up one of those genes (CRBP1) in cohorts of human ovarian cancer samples, using expression analysis in conjunction with laser capture microdissection and backed up by IHC analysis, and found that the expression pattern of CRBP1 was uniform in a tumor, either there was a complete loss in all cells in a tumor of CRBP1 or all the cells expressed the protein. Therefore I am curious if intratumor heterogeneity is dependent on the cell lineage and evolution of the transformed cell into a full tumor or a function of a discrete population of stem cells with varied genomic instability. Your results might suggest a more clonal evolution rather than a branched evolution which was found in this paper.
It is interesting that you mention the tough trials with the PTEN/PI3K/AKT axis of inhibitors. In high grade serous ovarian cancer we were never able to find any PI3K, PTEN, nor AKT mutations yet PI3K activity is usually overactive. If feel both your and Mohammed’s assessment that a systems biology approach instead of just relying on DNA mutational analysis will be more important in the future. In addition, there is nice work from Dr. Jefferey Peterson at Fox Chase and the development of a database of kinase inhibitors and activity effects on the kinome, showing the vast amount of crosstalk between once thought linear enzyme systems. If TKI’s will be the brunt of pharma’s development I feel they need to quickly develop as many TKI’s as they can now before we get to a clinical problem (resistance and lack of available therapeutics).

Emanuel Petricoin • Thanks Steven- yes, we are working with Charlie Swanton and Marco on the renal sets- our other studies are from breast and colon cancers. I think one of the things we do that really no one else is doing, unfortunately, is to laser capture microdissect the tumor cells from these specimens so that we have a more pure and accurate view of the signaling architecture. One confounder from the proteomic stand-point is the fact that pre-analytical variables such as post-excision delay times where the tissue is a hypoxic wound and signaling changes fluctuating as the tissue reacts to the ex-vivo condition can really effect things. When we look at tissue sets where the tissue is biopsied and immediately frozen we really dont see big differences in the signaling – the within tumor architecture is much more similar then between. We use the reverse phase array technology we invented to provide quantitative analysis on hundreds of phosphoproteins at once – so a nice view of the functional protein activation network. Your results of CRBP1 in ovarian tumors and the IHC data are very interesting. We will see how this all plays out. Of course once other confounder with the mutational data is that we really dont know what are the drivers and what are the passengers…
Yes I know Jeff Peterson’s work- its fantastic. In the end the hope I think- and in my personal opinion- will be rationally combined therapeutics based on the signaling architecture of each individual patient.

Incidentally, we just published a paper that you may be interested in from a “systems biology” standpoint-


Federici G, Gao X, Slawek J, Arodz T, Shitaye A, Wulfkuhle JD, De Maria R, Liotta LA, Petricoin EF 3rd. Mol Cancer Res. 2013 May

also- we published a paper that speaks directly to your point where we compared the signaling network activation of patient-matched primary colorectal cancers and synchronous liver mets. indeed there is huge systemic differences in the liver metastasis compared to the primary. there is no doubt in my mind that we will need to biopsy the metastasis to know how to treat. Looking at the primary tumor as a guide for therapy is a fools errand. here is the paper reference:

Protein pathway activation mapping of colorectal metastatic progression reveals metastasis-specific network alterations.

Silvestri A, Calvert V, Belluco C, Lipsky M, De Maria R, Deng J, Colombatti A, De Marchi F, Nitti D, Mammano E, Liotta L, Petricoin E, Pierobon M.

Clin Exp Metastasis. 2013 Mar;30(3):309-16. doi: 10.1007/s10585-012-9538-5. Epub 2012 Sep 29.

Center for Applied Proteomics and Molecular Medicine, George Mason University, 10900 University Blvd., Manassas, VA, 20110, USA.


The mechanism by which tissue microecology influences invasion and metastasis is largely unknown. Recent studies have indicated differences in the molecular architecture of the metastatic lesion compared to the primary tumor, however, systemic analysis of the alterations within the activated protein signaling network has not been described. Using laser capture microdissection, protein microarray technology, and a unique specimen collection of 34 matched primary colorectal cancers (CRC) and synchronous hepatic metastasis, the quantitative measurement of the total and activated/phosphorylated levels of 86 key signaling proteins was performed. Activation of the EGFR-PDGFR-cKIT network, in addition to PI3K/AKT pathway, was found uniquely activated in the hepatic metastatic lesions compared to the matched primary tumors. If validated in larger study sets, these findings may have potential clinical relevance since many of these activated signaling proteins are current targets for molecularly targeted therapeutics. Thus, these findings could lead to liver metastasis specific molecular therapies for CRC.

Adrian Anghel • I think both patterns (protein phosphorylation and mRNA) should be important in this complicated equation of heterogeneity. Let’s not forget the so-called functional miRNA-mRNA regulatory modules (FMRMs). Also I think we have different patterns of this heterogeneity for different evolutive stages of the tumour.


Alvin L. Beers, Jr., M.D. • This is a great study, but bad news for attempting to tailor treatment based on molecular markers. Dr. Swanton’s comment: “herterogeneity is likely to complicate matters” is an understatement. Intratumoral heterogeneity, branched, instead of linear, evolution of mutational events portends a nightmare in trying to predict location and volume of biopsies. I am reminded of a series of articles in Nature 491 (22 November 2012) “Physical Scientists take on Cancer”. There is a great comment by Jennie Dusheck: “Cancer researchers now recognize that taming wild cancer cells – populations of cells that evolve, cooperate, and roam freely through the body-demand a wider-angle view than molecular biology has been able to offer. Cross-disciplinary collaborations can approach cancer a greater spatial and temporal scales, using mathematical methods more typical of engineering, physics, ecology and evolutionary biology. The sense of failure so evident five years ago is giving way to the excitement of a productive intellectual partnership.” I’m not certain how well the “productive partnership” is going, but this Swanton study confirms the limitations of molecular biology.

Stephen J. Williams, Ph.D. • Thanks Dr. Beers for adding in your comment and adding in Jennie’s comment. Certainly it is something to be aware of if a cancer center’s strategy is to rely solely on gene arrays to genotype tumors. I think Dr. Pettricoin’s work on using proteomics might give some resolution to the matter however, in communicating with Dr. Swanton, I did not get the feeling of an “all hope is lost” but just that, in the case of solid tumors like renal, that careful monitoring of tumors after treatment may be warranted and, more interestingly, from a scientific standpoint, is the genetic complexity surrounding the origin of the disease, and not simple mutational spectrum of a single clone.

Burke Lillian • This is clinically a very important issue. Right now, sequencing or massive approaches such as pan-phosphorylation studies are helpful because, although we know many of the drivers, these studies are actually identifying new genes or new pathways that are activated. After a few (or several years), we truly will know which genes are typically activated and there will be panels to look for these.

Emanuel Petricoin • yes, I agree. In fact, the company that I co-founded, Theranostics Health, Inc– is launching a CLIA based protein pathway activation mapping test at ASCO that measures actionable drug targets (e.g. phospho HER2, EGFR, HER3, AKT, ERK, JAK, STAT, p70S6) and total HER2, EGFR, HER3 and PTEN. So these tests are coming even now.


Alvin L. Beers, Jr., M.D. • I do not think that “all hope is lost” nor did I have the impression that Dr. Swanton feels that way with regards to molecular profiling of cancer. I certainly applaud further research into the molecular aspects of cancer biology. But I do not believe that this will be sufficient. Integrating physicial sciences into cancer biology makes perfect sense toward better understanding of this complex disease.

Eleni Papadopoulos-Bergquist • I have enjoyed reading these comments and different ideas regarding genetic testing and profiling. As a nurse and researcher at heart, this is information that will make a huge impact on drug protocols, therefore allowing the best and most specific treatment to each individual rather than having a standard treatment protocol. Even with the scientific complexity of specifying genotypes of particular cancers, there is still the question of each individuals body responding to treatment. I’d love to have some dialogue regarding immune response.

Bradford Graves • I too have enjoyed reading this discussion. I am not a clinician but as a drug discovery researcher I have been struck by some parallels to the concept of virus fitness in virology – particularly as applied to HIV. Drug discovery cannot wait for the final answers to the many important questions being addressed in the discussion initiated by Dr. Williams. The best we can do is to pursue a broad range of therapeutics that will give the clinicians the armament they will need to either cure a given cancer or to at least turn it into a chronic as opposed to an acute disease. There has been a measure of success in the HIV field and it seems like it will be achievable for cancer. Obviously, to the extent that the labels of driver and passenger mutations can be correctly applied will help to prioritize the targets we address.

David W. Anderson • I would suggest that you look at the following publications:

Horn and Pao, (2009) JCO 26: 4232-4234.

Bunn and Doebele (2011) JCO:29:1-3

Boguski et al. (2009) Customized care 2020: how medical sequencing and network biology will enable personalized medicine. F1000 Bio Report 1:7.

Jones, S et al. (2010). Evolution of an adenocarcinoma in response to selection by targeted kinase inhibitors. Genome Biology. 11:R82. Marco Marra’s group in Toronto.

Also look at how companies and organizations like Foundation Medicine, Caris, Clarient, and CollabRx who are using genomics and sequencing on a large scale to address cancer from a personalized/individual approach.

Cancer is/will be a chronic disease requiring individualized/combinatorial therapies in many cases.

Alvin L. Beers, Jr., M.D. • David. These are excellent articles by Paul Bunn and Mark Boguski regarding integrating molecular markers into diagnostic evaluation, and I’ve seen other papers of similiar elk, and likely there will be more to come. Particularly in NSC lung cancer, the SOC is to use these markers up front. Diagnosis based on histology alone can no longer be recommended. The challenge for the future is how to integrate other aspects of cell biology with these markers. It remains daunting that not only do we see heterogeneity in molecular within tumors at a particularly point in time, but that there is often an evolution of markers over time, ie, a “plasticity” of markers, whether treatment is given or not. We know that targeted agents, TKI’s, enzyme inhibitors are not curative, but do give an improvement in PFS. A great deal of this resistance has to do with this “moving target” aspect of cancer cell biology..



1.         Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E, Martinez P, Matthews N, Stewart A, Tarpey P et al: Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. The New England journal of medicine 2012, 366(10):883-892.

2.         Caldas C: Cancer sequencing unravels clonal evolution. Nature biotechnology 2012, 30(5):408-410.

3.         Losi L, Baisse B, Bouzourene H, Benhattar J: Evolution of intratumoral genetic heterogeneity during colorectal cancer progression. Carcinogenesis 2005, 26(5):916-922.

4.         Krivtsov AV, Armstrong SA: Cancer. Can one cell influence cancer heterogeneity? Science 2012, 338(6110):1035-1036.

5.         Friedmann-Morvinski D, Bushong EA, Ke E, Soda Y, Marumoto T, Singer O, Ellisman MH, Verma IM: Dedifferentiation of neurons and astrocytes by oncogenes can induce gliomas in mice. Science 2012, 338(6110):1080-1084.

6.         Mintz B, Cronmiller C: Normal blood cells of anemic genotype in teratocarcinoma-derived mosaic mice. Proceedings of the National Academy of Sciences of the United States of America 1978, 75(12):6247-6251.

7.         Watanabe T, Dewey MJ, Mintz B: Teratocarcinoma cells as vehicles for introducing specific mutant mitochondrial genes into mice. Proceedings of the National Academy of Sciences of the United States of America 1978, 75(10):5113-5117.

8.         Mintz B, Cronmiller C, Custer RP: Somatic cell origin of teratocarcinomas. Proceedings of the National Academy of Sciences of the United States of America 1978, 75(6):2834-2838.



Other articles on this site on “PERSONALIZED MEDICINE” and “CANCER” and “OMICS” include:

Personalized medicine-based diagnostic test for NSCLC

Personalized medicine and Colon cancer

Helping Physicians identify Gene-Drug Interactions for Treatment Decisions: New ‘CLIPMERGE’ program – Personalized Medicine @ The Mount Sinai Medical Center

Systems Diagnostics – Real Personalized Medicine: David de Graaf, PhD, CEO, Selventa Inc.

Issues in Personalized Medicine in Cancer: Intratumor Heterogeneity and Branched Evolution Revealed by Multiregion Sequencing

Personalized Medicine: Clinical Aspiration of Microarrays

Understanding the Role of Personalized Medicine

Directions for Genomics in Personalized Medicine

Paradigm Shift in Human Genomics – Predictive Biomarkers and Personalized Medicine – Part 1

Rewriting the Mathematics of Tumor Growth; Teams Use Math Models to Sort Drivers from Passengers

Diagnosing Diseases & Gene Therapy: Precision Genome Editing and Cost-effective microRNA Profiling

Breast Cancer: Genomic profiling to predict Survival: Combination of Histopathology and Gene Expression Analysis

Proteomics and Biomarker Discovery


 Also please see our upcoming e-book “Genomics Orientations for Individualized Medicine” in our Medical E-book Series at












Read Full Post »

Personalized Medicine: Clinical Aspiration of Microarrays

Reporter, Writer: Stephen J. Williams, Ph.D.

 In this month’s Science, Mike May (at describes some of the challenges and successes in introducing microarray analysis to the clinical setting.  Traditionally used for investigational research, microarray is now being developed, customized and used for biomarker analysis, prognostic and predictive value, in a disease-specific manner.

Challenges in data interpretation

      In an interview with Seth Crosby, director of the Genome Technology Access Center at Washington University School of Medicine in St. Louis, “the biggest challenge” in moving microarray to the clinical setting is data interpretation.  The current technology makes it possible to evaluate expression of thousands of genes from a patient’s sample however as Crosby describes is assigning clinical relevance to the data.  For example Crosby explains that Washington University had validated a panel of 45 oncology genes by next generation sequencing and are using these genes to develop diagnostic tests to screen patient tumors for the purpose of determining a personalized therapeutic strategy. Seth Crosby noted it took “hundreds of Ph.D. and M.D. hours” to sift through the hundreds of papers to determine which genes were relevant to a specific cancer type. However, he notes, that once we better understand which changes in the patient’s genome are related to a specific disease we will be able to narrow down the list and be able to produce both economical and more disease-relevant microarrays.

Is this aberration pathogenic or not?

     Microarrays are becoming an invaluable tool in cytogenetics, as eluded by Andy Last, executive vice president of the genetic analysis business unit at Affymetrix.  Certain diseases like Down syndrome have well characterized chromosomal alterations like additions or deletions of parts or entire chromosomes.  According to Affymetrix, the most common use of microarrays is for determining copy number variation.  However according to James Clough, vice president of clinical and genomic services at Oxford Gene Technology, given the hundreds of syndromes associated with chromosomal rearrangements, the challenge will be to determine if a small chromosomal aberration has pathologic significance, given that microarray affords much higher diagnostic yield and speed of analysis than traditional microscopic techniques.  To address this challenge, Oxford Gene Technologies, PerkinElmer, Affymetrix, and Agilent all have custom designed microarrays to evaluate disease specific copy number and SNP (single nucleotide polymorphism) microarrays.  For example PerkinElmer designed OncoChip™ to evaluate copy number variation in more than 1.800 cancer genes.  Agilent makes microarrays that evaluates both copy number variation such as its CGH (comparative genomic hybridization) plus SNP microarrays.  Patricia Barco, product manager for cytogenetics at Agilent, notes these arrays can be used in prenatal and postnatal research and cancer, and “can be customized from more than 28 million probes in our library”.

Custom Tools and Software to Handle the Onslaught of Big Data

     There is a need for FDA approved diagnostic tools based on microarrays. Pathwork Diagnostic’s has one such tool (the Pathwork Tissue of Origin test), which uses 2,000 transcript markers and a proprietary computational algorithm to determine from expression analysis, the tissue of origin of a patient’s tumor.  Pathwork also provides a fast, custom turn-around analytical service for pathologists who encounter difficult to interpret samples.  Illumina provides the Infinium HumanCore BeadChip family of microarrays, which can determine genetic variations for purposes of biological tissue banking.  This system uses a set of over 300,000 SNP probes plus 240,000 exome-based markers.

     Tools have also been developed to validate microarray results.  A common validation strategy is the use of quantitative real-time PCR to verify the expression changes seen on the microarray.  Life Technologies developed the TaqMan OpenArray Real Time PCR plates, which have 3,072 wells and can be custom-formatted using their library of eight million validated TaqMan assays.

Making Sense of the Big Data: Bridging the Knowledge Gap using Bioinformatics

          The use of microarray has spurned industries devoted to developing the bioinformatics software to analyze the massive amounts of data and provide clinical significance.  For example companies such as Expression Analysis use their bioinformatics software to provide pathway analysis for microarray data in order to translate the data into the biology.  Using such strategies can also validate the design of microarrays for various diseases.

Foundation Medicine, Inc., a molecular information company, provides cancer genomics test solutions. It offers FoundationOne, an informative genomic profile to identify a patient’s individual molecular alterations and match them with relevant targeted therapies and clinical trials. The company’s product enables physicians to recommend treatment options for patients based on the molecular subtype of their cancer.

The Canadian Bioinformatics Workshops series recently offered a course on using bioinformatic approaches to analyze clinical data generated from microarray approaches (   The course objectives are described below:

Course Objectives

Cancer research has rapidly embraced high throughput technologies into its research, using various microarray, tissue array, and next generation sequencing platforms. The result has been a rapid increase in cancer data output and data types. Now more than ever, having the bioinformatic skills and knowledge of available bioinformatic resources specific to cancer is critical. The CBW will host a 5-day workshop covering the key bioinformatics concepts and tools required to analyze cancer genomic data sets. Participants will gain experience in genomic data visualization tools which will be applied throughout the development of the skills required to analyze cancer -omic data for gene expression, genome rearrangement, somatic mutations and copy number variation. The workshop will conclude with analyzing and conducting pathway analysis on the resultant cancer gene list and integration of clinical data.

Successful Examples of Clinical Ventures Integrating Bioinformatics in Cancer Treatment Decision –Making

The University of Pavia, Italy developed a fully integrated oncology bioinformatics workflow as described on their website and at the ESMO 2012 Congress meeting:






ESMO 2012




Translational research


A. Zambelli, D. Segagni, V. Tibollo, A. Dagliati, A. Malovini, V. Fotia, S. Manera, R. Bellazzi; Pavia/IT

  • Body

The ONCO-i2b2 project, supported by the University of Pavia and the Fondazione Salvatore Maugeri (FSM), aims at supporting translational research in oncology and exploits the software solutions implemented by the Informatics for Integrating Biology and the Bedside (i2b2) research centre, an initiative funded by the NIH Roadmap National Centres for Biomedical Computing. The ONCO-i2b2 software is designed to integrate the i2b2 infrastructure with the FSM hospital information system and the Bruno Boerci Biobank, in order to provide well-characterized cancer specimens along with an accurate patients clinical data-base. The i2b2 infrastructure provides a web-based access to all the electronic medical records of cancer patients, and allow researchers analyzing the vast amount of biological and clinical information, relying on a user-friendly interface. Data coming from multiple sources are integrated and jointly queried.

In 2011 at AIOM Meeting we reported the preliminary experience of the ONCO-i2b2 project, now we’re able to present the up and running platform and the extended data set. Currently, more than 4400 specimens are stored and more than 600 of breast cancer patients give the consent for the use of specimens in the context of clinical research, in addition, more than 5000 histological reports are stored in order to integrate clinical data.

Within the ONCO-i2b2 project is possible to query and merge data regarding:

• Anonymous patient personal data;

• Diagnosis and therapy ICD9-CM subset from the hospital information system;

• Histological data (tumour SNOMED and TNM codes) and receptor profile testing (Her2, Ki67) from anatomic pathology database;

• Specimen molecular characteristics (DNA, RNA, blood, plasma and cancer tissues) from the Bruno Boerci Biobank management system.

The research infrastructure will be completed by the development of new set of components designed to enhance the ability of an i2b2 hive to utilize data generated by NGS technology, providing a mechanism to apply custom genomic annotations. The translational tool created at FSM is a concrete example regarding how the integration of different information from heterogeneous sources could bring scientific research closer to understand the nature of disease itself and to create novel diagnostics through handy interfaces.


All authors have declared no conflicts of interest.

NCI has under-taken a similar effort under the Recovery Act (the full text of the latest report is taken from their website

Cancer Bioinformatics: Recovery Act Investment Report

November 2009

Public Health Burden of Cancer

Cancer is the second leading cause of death in the United States after heart disease. In 2009, it is estimated that nearly 1.5 million new cases of invasive cancer will be diagnosed in this country and more than 560,000 people will die of the disease.

To learn more, visit:

Cancer Bioinformatics Program Overview

Over the past five years, NCI’s Center for Biomedical Informatics and Information Technology (CBIIT) has led the effort to develop and deploy the cancer Biomedical Informatics Grid® (caBIG) in partnership with the broader cancer community.  The caBIG network is designed to enable the integration and exchange of data among researchers in the laboratory and the clinic, simplify collaboration, and realize the potential of information-based (personalized) medicine in improving patient outcomes. caBIG has connected major components of the cancer community, including NCI-designated Cancer Centers, participating institutions of the NCI Community Cancer Centers Program (NCCCP), and numerous large-scale scientific endeavors, as well as basic, translational, and clinical researchers at public and private institutions across the United States and around the world.  Beyond cancer research, caBIG capabilities—infrastructure, standards, and tools—provide a prototype for linking other disease communities and catalyzing a new 21st-century biomedical ecosystem that unifies research and care. ARRA funding will allow NCI to accelerate the ongoing development of the Cancer Knowledge Cloud and Oncology Electronic Health Records (EHRs) initiatives, thereby providing for continued job creation in the areas of biomedical informatics development and application as well as healthcare delivery.

The caBIG Cancer Knowledge Cloud: Extending the Research Infrastructure

The Cancer Knowledge Cloud is a virtual biomedical capability that utilizes caBIG tools, infrastructure, and security frameworks to integrate distributed individual and organizational data, software applications, and computational capacity throughout the broad cancer research and treatment community. The Cancer Knowledge Cloud connects, integrates, and facilitates sharing of the diverse primary data generated through basic and clinical research and care delivery to enable personalized medicine. The cloud includes information generated through large-scale research projects such as The Cancer Genome Atlas (TCGA), the cancer Human Biobank (caHUB) tissue acquisition network, the NCI Functional Biology Consortium, the NCI Patient Characterization Center, and the NCI Preclinical Development Pipeline, academic and industry counterparts to these projects, and clinical observations (from entities such as the NCCCP) captured in oncology-extended Electronic Health Records.  Through the use of the caBIG Data Sharing and Security Framework, the Cloud will support appropriate sharing of information, supporting in silico hypothesis generation and testing, and enabling a learning healthcare system.

A caBIG-Based Rapid-Learning Healthcare System: Incorporating Oncology-Extended Electronic Healthcare Records (EHRs)

The 21st-century Cancer Knowledge Cloud will connect individuals, organizations, institutions, and their associated information within an information technology-enabled cycle of discovery, development, and clinical care—the paradigm of a rapid-learning healthcare system. This will transform these disconnected sectors into a system that is personalized, preventive, pre-emptive, and patient-participatory.  To be realized, this model requires the adoption of standards-based EHRs. Presently, however, no certified oncology-based EHR exists, and fewer than 3 percent of oncologists with outpatient-based practices utilize EHRs. caBIG has recently established a collaboration with the American Society of Clinical Oncology (ASCO) to develop an oncology-specific EHR (caEHR) specification based on open standards already in use in the oncology community that will utilize caBIG standards for interoperability. NCI will implement an open-source version of this specification to validate the specification and to provide a free alternative to sites that choose not to purchase a commercial system. The launch customer for the caEHR will be NCCCP participating sites. NCI will work with appropriate entities to provide a mechanism for certifying that caEHR implementations are consistent with the NCI/ASCO specification.

Bards Cancer Institute has another clinical bioinformatics program to support their clinical efforts:

Clinical Bioinformatics Program in Oncology at Barts Cancer Institute at Barts and the London School of Medicine

BCI HomeCancer Bioinformatics


Why we focus on Cancer Bioinformatics

Bioinformatics is a new interdisciplinary area involving biological, statistical and computational sciences. Bioinformatics will enable cancer researchers not only to manage, analyze, mine and understand the currently accumulated, valuable, high-throughput data, but also to integrate these in their current research programs. The need for bioinformatics will become ever more important as new technologies increase the already exponential rate at which cancer data are generated.

What we do

  • We work alongside clinical and basic scientists to support the cancer projects within BCI.  This is an ideal partnership between scientific experts, who know the research questions that will be relevant from a cancer biologist or clinician’s perspective, and bioinformatics experts, who know how to develop the proposed methods to provide answers.
  • We also conduct independent bioinformatics research, focusing on the development of computational and integrative methods, algorithms, databases and tools to tackle the analysis of the high volumes of cancer data.
  • We also are actively involved in the development of bioinformatics educational courses at BCI. Our courses offer a unique opportunity for biologists to gain a basic understanding in the use of bioinformatics methods to access and harness large complicated high-throughput data and uncover meaningful information that could be used to understand molecular mechanisms and develop novel targeted therapeutics/diagnostic tools.

Developing Criteria for Genomic Profiling in Lung Cancer:

A Report from U.S. Cancer Centers

In a report by Pao et. al., a group of clinicians organized a meeting to standardize some protocols for the integration of microarray and genomic data from lung cancer patients into the clinical setting.[1]  There has been ample evidence that adenocarcinomas could be classified into “clinically relevant molecular subsets” based on distinct genomic changes.  For example EGFR (epidermal growth factor receptor) exon 19 deletions and exon 21 point mutations predict sensitivity to tyrosine kinase inhibitors (TKIs) like gefitinib, whereas exon 20 insertions predict primary resistance[2].

However, as the authors note, “mutational profiling has not been widely accepted or adopted into practice in thoracic oncology”.  

     Therefore, a multi-institutional workshop was held in 2009 among participants from Massachusetts General Hospital (MGH) Cancer Center, Memorial Sloan-Kettering Cancer Center (MSKCC), the Dana-Farber/Bingham & Women’s Cancer Center (DF/BWCC), the M.D. Anderson Cancer Center (VICC), and the Vanderbilt-Ingram Cancer Center (VICC) to discuss their institutes molecular profiling programs with emphasis on:

·         Organization/workflow

·         Mutation detection technologies

·         Clinical protocols and reporting

·         Patient consent

In addition to the aforementioned challenges, the panel discussed further issues for developing improved science-driven criteria for determining targeted therapies including:

1)      Including pathologists into criteria development as pathology departments are usually the main repositories for specimens

2)      Developing integrated informatics systems

3)      Standardizing new target validation methodology across cancer centers


1.            Pao W, Kris MG, Iafrate AJ, Ladanyi M, Janne PA, Wistuba, II, Miake-Lye R, Herbst RS, Carbone DP, Johnson BE et al: Integration of molecular profiling into the lung cancer clinic. Clinical cancer research : an official journal of the American Association for Cancer Research 2009, 15(17):5317-5322.

2.            Wu JY, Wu SG, Yang CH, Gow CH, Chang YL, Yu CJ, Shih JY, Yang PC: Lung cancer with epidermal growth factor receptor exon 20 mutations is associated with poor gefitinib treatment response. Clinical cancer research : an official journal of the American Association for Cancer Research 2008, 14(15):4877-4882.

Other posts on this website on Cancer and Genomics include:

Read Full Post »

A perspective on where we are on carcinogenesis, cancer variability and predictors of time to recurrence and future behavior

Author: Larry H Bernstein, MD, FCAP

I.     Background

In “Tumor Imaging and Targeting: Predicting Tumor Response to Treatment: Where we stand? “ ( Dec 13, 2012) Dr. Ritu Saxena  attempts to integrate three posts and to embed all comments made to all three papers, allowing the reader a critically thought compilation of evidence-based medicine and scientific discourse.

Dr. Dror Nir authored a post on October 16th titled “Knowing the tumor’s size and location, could we target treatment to THE ROI by applying imaging-guided intervention?” The article attracted over 20 comments from readers including researchers and oncologists debating the following issues:

imaging technologies in cancer

  • tumor size, and
  • tumor response to treatment.

The debate lead to several new posts authored by:

Dr. Bernstein’s (What can we expect of tumor therapeutic response),

Dr. Saxena, the Author of this post’s, (Judging ‘tumor response’-there is more food for thought) and

Dr. Lev-Ari’s post on Personalized Medicine: Cancer Cell Biology and Minimally Invasive Surgery (MIS)

The post was a compilation of the views of authors representing different specialties including research and medicine. In medicine: Pathology, Oncology Surgery and Medical Imaging, are represented.

Dror Nir added a fresh discussion in “New clinical results supports Imaging-guidance for targeted prostate biopsy” based on a study of “Artemis”, a system that is adjunct to ultrasound and performs 3D Imaging and Navigation for Prostate Biopsy by Eigen (a complementary post to “Imaging-guided biopsies: Is there a preferred strategy to choose?”).

Image fusion is the process of combining multiple images from various sources into a single representative image. Ultrasound is the imaging modality used to guide Artemis in performing the biopsies. In this study MRI is used to overcome the “blindness” regarding tumor location.  This supports the detection reliability issue made in his “ Imaging-guided biopsies: Is there a preferred strategy to choose?” and  “Fundamental challenge in Prostate cancer screening.”

This makes the case that In the future, MRI-ultrasound fusion for lesion targeting is likely to result in fewer and more accurate prostate biopsies than the present use of systematic biopsies with ultrasound guidance alone.   Nevertheless, we haven’t completed the case for prediction of recurrence, even if we may eliminate the unnecessary consequences of radical prostatectomy.

Let’s look a little further. A discussion opens up more questions for discussion. I just read an interesting related article. The door has opened  wider.

II.               Novel technology to detect cancer in early stages

A. nanoparticles

Researchers have developed novel technology to detect the tumors in the body in early stages with the help of nanoparticles .( Nature Biotechnology).

Cancer cells produce many of the proteins that could be used as biomarkers to detect the cancer in the body but the amount of these proteins is not up to the mark or they may get diluted in the body of the patients making it nearly impossible to detect them in early stages.

This new technology has been developed by the researchers from MIT . Nanoparticles (brown) coated with peptides (blue) cleaved by enzymes (green) at the disease site. Peptides than come into the urine to be detected by mass spectrometry. (Credit: Justin H. Lo/MIT)

In this technology, nanoparticles will interact with the tumor proteins helping to make thousands of biomarkers secreted by the cancer cells. We had this ‘aha’ moment: What if you could deliver something that could amplify the signal?”

  • Scientists administered ‘synthetic biomarkers’ having peptides bonded to the nanoparticles and
  • the particles interact with the protease enzymes often found in large quantities in cancer cells

as they help them to cut the proteins normally holding the cells in place and to spread in other parts of the body.

Researchers found that the proteases break down hundreds of peptides from the nanoparticles and release them in the bloodstream. These peptides are then excreted in the urine, where the process of mass spectrometry could help to detect such peptides.

These “Synthetic biomarkers” perform three functions in vivo:

  1. they target sites of disease,
  2. sample dysregulated protease activities and
  3. emit mass-encoded reporters into host urine (for multiplexed detection by MS).

According to Bhatia, this biomarker amplification technology could also be used to manage the advancement of the disease and to check the response of the tumors to the drugs.


Kwong, G., von Maltzahn, G., Murugappan, G., Abudayyeh, O., Mo, S., Papayannopoulos, I., Sverdlov, D., Liu, S., Warren, A., Popov, Y., Schuppan, D., & Bhatia, S. (2012). Mass-encoded synthetic biomarkers for multiplexed urinary monitoring of disease Nature Biotechnology

IIB    Synthetic Nucleosides

J Gong and SJ Sturla published “A Synthetic Nucleoside Probe that Discerns a DNA Adduct from Unmodified DNA” in JACS Communications on web 4/03/2007).  They state that biologically reactive chemicals alkylate DNA and induce structural modifications in the form of covalent adducts that can persist, escape repair, and serve as templates for polymerase-mediated DNA synthesis. Therefore, correlating chemical structures and quantitative levels of adducts with toxicity is essential for targeting specific agents to carcinogenesis.

  • DNA adducts are formed at exceedingly low levels.
  • Minor lesions may have greater biological impact than more abundant products.
  • New molecular approaches for addressing specific low-abundance adducts are needed

They describe the first example of a synthetic nucleoside that may serve as the chemical basis for a probe of a bulky carcinogen-DNA adduct

IIC.  MicroRNAs caused by DNA methylation

Another molecular approach “ A microRNA DNA methylation signature for human cancer metastasis” was published in PNAS [2008;105(36):13556-13561)] by A Lujambio , Calin GA, Villanueva A et al.

Different sets of miRNAs are usually deregulated in different cancers, and some miRNAs are aberrantly methylated and silenced, causing tumorigenesis. The authors

  • identified aberrantly methylated and silenced miRNAs that are cancer-specific
  • using miRNA microarray techniques.

Functional analyses for the selected genes proved that these miRNAs act on C-MYC, E2F3, CDK6 and TGIF2, resulting in metastasis through aberrant methylation of the miRNAs. The authors suggest that these may be applicable to advance research in the clinical setting.

III.              New methods require advanced mathematical prediction methods

A.  First Case …ProsVue PSA

One of the most elegant papers I have seen in several years  has been published in Clinical Biochemistry (CLB–12-00159), by Mark J. Sarnoa1 and Charles S. Davis2. [1Vision Biotechnology Consulting, 19833 Fortuna Del Este Road, Escondido, CA 92029, USA (, 2CSD Biostatistics, Inc., San Diego, CA, 4860 Barlows Landing Cove, San Diego, CA 92130, USA (]

Robustness of ProsVue™ linear slope for prediction of prostate cancer recurrence: Simulation studies on effects of analytical imprecision and sampling time variation.
Keywords: ProsVue, slope, prostate cancer, random variates.
Financial support for the investigation was provided by Iris Molecular Diagnostics

Abstract: Objective: The ProsVue assay measures

  • serum total prostate-specific antigen (PSA) over three time points post-radical prostatectomy and
  • calculates rate of change expressed as linear slope. Slopes ≤2.0 pg/ml/month are associated with reduced risk for prostate cancer recurrence.

However, an indicator based on measurement at multiple time points, calculation of slope, and relation of slope to a binary cutpoint may be subject to effects of analytical imprecision and sampling time variation.

They performed simulation studies to determine the presence and magnitude of such effects.

Design and Methods: Using data from a two-site precision study and a multicenter retrospective clinical trial of 304 men, they carried out simulation studies to assess whether analytical imprecision and sampling time variation can drive misclassification of patients with stable disease or classification switching for patients with clinical recurrence.


  • Analytical imprecision related to expected PSA values in a stable disease population results in ≤1.2% misclassifications.
  • For recurrent populations, an analysis taking into account correlation between sampling time points demonstrated that classification switching across the 2.0 pg/ml/month cutpoint occurs at a rate ≤11%.
  • Lastly, sampling time variation across a wide range of scenarios results in 99.7% retention of proper classification for stable disease patients with linear slopes up to the 75th percentile of the distribution.


  • These results demonstrate the robustness of the ProsVue assay and the linear slope indicator.
  • Further, these simulation studies provide a potential framework for evaluation of future assays that may rely on the rate of change principle

The ProsVue Assay has been cleared for commercial use by the US Food and Drug Administration (FDA) as “a prognostic marker in conjunction with clinical evaluation as an aid in

  • identifying those patients at reduced risk for recurrence of prostate cancer for the eight year period following prostatectomy.”

The assay measures

  1. serum total prostate specific antigen (PSA) in post-RP samples and
  2. calculates rate of change of PSA over the sampling period,

expressing the outcome as linear slope. The assay is novel in at least a few respects.

  • the assay is optimized to identify patients at reduced risk for recurrence.

In order to demonstrate efficacy for this indication, the assay employs the immuno-polymerase chain reaction (immuno-PCR) to achieve sensitivity

  • an order of magnitude lower than existing “ultrasensitive” PSA assays.

The improved sensitivity allows quantification of PSA at levels exhibited in stable disease (<5 pg/ml), which have been historically below the

measurement range of ultrasensitive assays.

Secondly, the assay is the first to receive clearance based on

  • linear slope of tumor marker concentration versus time post-surgery.
  • Specifically, PSA is measured in three samples taken between 1.5 and 20 months post-RP and
  • the slope calculated using simple least squares regression.
  • The calculated slope is compared to a threshold of 2.0 pg/ml/month with values at or below the threshold associated with reduced risk for PCa recurrence.

Does analytical imprecision present a potential risk for misclassification by driving errors in the calculated slope that result in classification switching?  Since excursions of precision can occur as point sources in single sampling points or in cumulative effect from the three sampling points, the question is worthy of consideration. They carried out studies

  • to address these questions specific to ProsVue and also
  • provide a potential framework for evaluation of future assays.
  • Similarly, does variation in the time at which samples are taken drive errors resulting in classification switching?

Both questions require evaluating the robustness of the ProsVue Assay and are properly presented for clinical chemists and physicians evaluating use of the assay in clinical practice. Furthermore, since future diagnostic assays may employ the rate of change principle, it is important to develop statistical methods to evaluate effects of variation.

The point is that more sophisticated methods are needed to measure scarce analytes associated with risk for eventual clinical events.

  • Accurate measurement at post-RP levels to identify patients with reduced risk of recurrence represents a new development.
  • Furthermore, measurement of PSA at multiple time points and calculation of rate of change using linear regression extends application of the analyte markedly beyond traditional use.

Such use presents certain questions of variation effects.

Their results indicate that analytical imprecision in the range of concentrations exhibited in patients at reduced risk for recurrence (the focus of the assay) presents no significant risk of misclassification.

  • Classification switching in this population occurs at a frequency of ≤1.2%.
  • Slopes for recurrent patients and clinical classification are substantively insensitive to analytical variation even in a subpopulation of recurrent patients with slowly rising PSA values.
  • Sampling time variation negligibly affects clinical classification for stable disease patients with slopes at and below the 75th percentile.
Table 1. Side-effects and effects on recovery ...

Table 1. Side-effects and effects on recovery of treatments for newly diagnosed prostate cancer. The Prostate Brachytherapy Advisory Group: (Photo credit: Wikipedia)


IIIB. Other interesting developments are going to need further development and validation.

For instance, research has been published online in the journal Cancer Cell, reports a cellular component that is involved in mobility of cancer to other body parts and inhibition of which could increase the tumor formation. These investigators worked on various animal models including chicken, zebrafish and mouse, and patient samples and have found a cellular component; Prrx1 that stops the cancer cells from staying in organs.  Epithelial-mesenchymal transition (EMT) is the process that is required by the cancer cells to spread to other organs. This process helps the cells to become mobile and move with the bloodstream. These cells must lose their mobility before attaching to other body parts.

In the final analysis the cells have to lose the component Prrx1 to lose mobility and to become stationary. Researchers wrote, “Prrx1 loss reverts EMT & induces stemness, both required for metastatic colonization.”  Consequently,  Prrx1 has to be turned off for these cells to group together to form other tumours.” It has been found that the tumors with elevated levels of Prrx1 cannot form new tumors.

IIIC.  PXR and AhR Nuclear Receptor Activation

  • The primary mechanism of cytochrome P450 induction is via increased gene transcription which typically occurs through nuclear receptor activation.
  • The most common nuclear receptors involved in the induction of drug metabolizing enzymes include the pregnane X receptor (PXR), the aryl hydrocarbon receptor (AhR), and the constitutive androstane receptor (CAR) which are known to regulate CYP3A4, CYP1A2 and CYP2B6, respectively.
  • An industry survey of current practices and recommendations (Chu et al., (2009) Drug Metab Dispos 37: 1339-1354) indicates 64% of survey respondents routinely use nuclear receptor transactivation assays to assess the potential of test compounds to cause enzyme induction
  • ‘Because reporter assays are relatively high throughput and cost effective, they can be a valuable tool in drug discovery.’(Chu V, Einolf HJ, Evers R, Kumar G, et.  (2009) Drug Metab Dispos 37; 1339-1354)
  • Luciferase reporter gene assay

No cytotoxicity was observed for any of six compounds at the concentration range tested with the exception of troglitazone for which cytotoxicity was observed at the highest concentration of 50μM.  This data point was excluded in this instance and not used for calculating the Emax or EC50.

In brief, CAR and PXR regulate distinct but overlapping sets of target genes, which include certain phase 1 P450 enzymes (e.g., CYP2B, CYP3A, and CYP2C), phase II conjugation enzymes such as UDP glucuronosyltransferase UGT1A1 and sulfotransferase SULT2A, and phase III transporters such as P-glycoprotein (MDR-1). The AhR receptor has been shown to regulate the expression of CYP1A.

Will this be combined with the other methods for drug selection and prediction of drug free survival?

I have mentioned an improved molecular assay of PSA at the pcg/ml level that is approved for use with an acceptable linear prediction of survival for 8 years post radical prostatectomy.  Then there is a report of a method of measuring nanoparticles in urine, to amplify the signal detected by mass spectrometry. This new technology has been developed by the researchers from MIT and led by Sangeeta Bhatia at MIT. (Novel technology to detect cancer in early stages, Nature Biotechnology, Dec 16, 2012).  There is still another recent report about using gene expression profiles to predict breast cancer, and a number of articles have shown variability in breast cancer types.   I view with reservations until I can see long term predictions of prognosis.

IIID.  Prediction of Breast Cancer Metastasis by Gene Expression Profiles

The report in Cancer Informatics (; open access)  by M Burton, M Thomassen, Q Tan, and TA Kruse is “Prediction of Breast Cancer Metastasis by Gene Expression Profiles: A Comparison of Metagenes and Single Genes.”   The authors state “The diversity of microarray platforms has made the full validation of gene expression  profiles across studies difficult and, the classification accuracies are rarely validated in multiple independent datasets. The individual genes between such lists may not match, but genes with comparable function are included across gene lists. However,  genes can be grouped together as metagenes (MGs) based on common characteristics such as pathways, regulation, or genomic location. Such MGs might be used as features in building a predictive model applicable for classifying independent data.”

Microarray gene expression analysis has in several previous studies been applied to elucidate the relation between clinical outcome and gene expression patterns in breast cancer and has demonstrated improvement of recurrence prediction. In some studies, genes in such profiles might be fully or partially missing in the test data used for validation due to the choice of microarray platform or the presence of missing values associated with a given probe.

To overcome the obstacles, these authors propose that individual genes could be considered part of a larger network such that their expression being controlled by the expression level of other genes or that a group of genes belong to a specific pathway performing a well-defined task. These genes may be controlled by the same transcription factor or located in the same chromosomal region. In fact these groupings have been collected in public databases (the Kyoto Encyclopedia of Genes and Genomes (KEGG), the Molecular Signature Database (MsigDB), the Gene Ontology database (GO)). This could be upregulation or deregulation of pathways associated with metastasis. Metastasis progressionas well as tumor grading (in breast cancer) are associated with accumulated mutations in several genes, leading to amplification or inactivation of genes.

Several studies have defined metagene/gene modules derived from microarray data using various methods such as penalized matrix decomposition which clusters similar genes but without similar expression profiles – hierarchical clustering, correlation, or combining a priori protein-protein interactions with microarray gene expression data defining interaction networks as features. Few studies have attempted to use such predefined gene sets for prediction models.

Their study compared the performance of either metagene- or single gene-based feature sets and classifiers using random forest and two support vector machines for classifier building. The performance within the same dataset, feature set validation performance, and validation performance of entire classifiers in strictly independent datasets were assessed by 10 times repeated 10-fold cross validation, leave-one-out cross validation, and one-fold validation, respectively. To test the significance of the performance difference between MG- and SG-features/classifiers, we used a repeated down-sampled binomial test approach.

They found MG- and SG-feature sets are transferable and perform well for training and testing prediction of metastasis outcome in strictly independent data sets, both between different and within similar microarray platforms.  Further, The study showed that MG- and SG-feature sets perform equally well in classifying independent data. Furthermore, SG-classifiers significantly outperformed MG-classifier when validation is conducted between datasets using similar platforms, while no significant performance difference was found when validation was performed between different platforms.

  • The MG- and SG-classifiers had similar performance when conducting classifier validation in independent data based on a different microarray platform.
  • The latter was also true when only validating sets of MG- and SG-features in independent datasets, both between and within similar and different platforms.

This study appears to be unique in the same way that the PCa prediction study is unique in that genome-based expression patterns are used to classify and predict metastatic potential.

These studies have the potential to materialize into practice changing behavior.

IIIE. Colon Cancer and Treatment Recurrence

Cancer scientists led by Dr. John Dick at the Princess Margaret Cancer Centre have found a way to follow single tumour cells and observe their growth over time. By using special immune-deficient mice to propagate human colorectal cancer, they found that genetic mutations, regarded by many as the chief suspect driving cancer growth, are only one piece of the puzzle. The team discovered that biological factors and cell behaviour — not only genes — drive tumour growth, contributing to therapy failure and relapse. The findings are published December 13 online ahead of print in Science, are “a major conceptual advance in understanding tumour growth and treatment response” according to Dr. Dick.

[1] only some cancer cells are responsible for keeping the cancer growing.

[2] these kept the cancer growing for long time periods (up to 500 days of repeated tumour transplantation)

[3] a class of propagating cancer cells that could lie dormant before being activated.

[4] the mutated cancer genes were identical for all of these different cell behaviours.

[5] given chemotherapy the long-term propagating cells were generally sensitive to treatment, but dormant cells were not killed by drug treatment.

[6] these became activated andpropagated new tumour.

IV. Related References

Diagnostic efficiency of carcinoembryonic antigen and CA125 in the cytological evaluation of effusions.
M M Pinto, L H Bernstein, R A Rudolph, D A Brogan, M Rosman
Arch Pathol Lab Med 1992; 116(6):626-631 ; ICID: 825503

Medically significant concentrations of prostate-specific antigen in serum assessed.
L H Bernstein, R A Rudolph, M M Pinto, N Viner, H Zuckerman
Clin Chem 1990; 36(3):515-518 ; ICID: 825497

Entropy and Information Content of Laboratory Test Results
R T Vollmer
Am J Clin Pathol.  2007;127(1):60-65.


This article introduces the use of information theoretic concepts such as entropy, S, for the evaluation of laboratory test results, and it offers a new measure of information, 1 – S,
which tells us just how far toward certainty a laboratory test result can predict a binary outcome. The derived method is applied to the serum markers troponin I and
prostate-specific antigen and to histologic grading of HER-2/neu staining, to cytologic diagnosis of cervical specimens, and to the measurement of tumor thickness in malignant
melanoma. Not only do the graphic results provide insight for these tests, they also validate prior conclusions. Thus, this information theoretic approach shows promise for
evaluating and understanding laboratory test results.

A map of protein-protein interactions involving calmodulin. Protein-protein interactions are both numerous and incredibly complex, and they can be mapped using the Database of
Interacting Proteins (DIP). This image depicts a DIP map for the protein calmodulin. The interactions with the most confidence are drawn with wider connecting lines. This diagram
highlights one level of complexity involved in understanding the downstream effects of gene regulation and expression.

Related article

Read Full Post »

Reporter and Curator: Dr. Sudipta Saha, Ph.D.

With the completion of the mapping of the human genome, we now have access to all the DNA sequence information responsible for human biology. Together with microarray technology, we are ushering in a new era in reproductive medicine—the era of Reproductive Genomics.

Whole genome microarray analysis of the testis and ovary suggests that a substantial part of the genome is expressed in reproductive tissues and many of them are likely to be important for normal reproduction. Yet adequate expression and functional information is only available for less than 10% of them. Hence, one of the important questions in reproductive studies now is ‘how do we associate function with the genes expressed in reproductive tissues?’ The establishment of mutations in animal models such as the mouse represents one powerful approach to address this question.

Animal models have played critical roles in improving our understanding of mechanisms and pathogenesis of diseases. Mouse knockout models have often provided highly needed functional validation of genes implicated in human diseases. The rapid advance of human genetics in areas such as

  • single nucleotide polymorphisms (SNP) and
  • haplotyping technology

now allows the identification of disease-associated single nucleotide variation at a much faster pace. Functional examination of those candidate genes is needed to determine if those genes or variants are indeed involved in reproductive disease. Generating mutations in murine homologs of candidate genes represents a direct way to determine their roles, and mouse models will further allow the dissection of genetic pathways underlying the disease condition and provide models to test possible drug treatments. Thus, how to generate mouse models efficiently becomes a priority issue in the Genomics era of Reproductive Medicine.

It is known that generating a mouse knockout is no small endeavor, even for a mouse research lab, often requiring specialized expertise and experience in

  • molecular biology,
  • embryonic stem (ES) biology and
  • mouse husbandry.

Therefore, it could be intimidating for people who have little experience in mouse research. Fortunately, there are some technological developments in the mouse community that make the task of generating mouse mutations less intimidating to people unfamiliar with mouse genetics. One of these developments is the effort led by the International Gene Trap Consortium (IGTC) to generate a library of mouse mutant ES cells covering most of the genes in the mouse genome. This method saves researchers the tedious and sometimes challenging tasks of making knockout vectors and screening ES cell colonies and directly provides researchers an ES cell clone carrying the mutation of the gene of interest.

Because gene trapping involves the use of different mechanisms in generating mutations from the traditional knockout method, and its efficacy in targeting reproductive genes which often are expressed in later development or adult has not been fully established, it is necessary to examine the benefits and limitations of this technology, especially in the perspective of reproductive medicine so that reproductive researchers and physicians who are interested in mouse models could become familiar with this technology.

With this in mind, we provide an overview of the gene trapping mutagenesis method and its possible application to Reproductive Medicine. We evaluate gene trapping as a method in terms of its efficiency in comparison with traditional knockout methods and use an in-house software program to screen the IGTC database for existing cell lines with possible mutations in genes expressed in various reproductive tissues. Among over seven thousand genes highly expressed in human ovaries, almost half of them have existing gene trap lines.

Additionally, from 900 human seminal fluid proteins, 43% of them have gene trap hits in their mouse homologs. Our analysis suggests gene trapping is an effective mutagenesis method for identifying the genetic basis of reproductive diseases and many mutations for important reproductive genes are already present in the database. Given the rapid growth of the number of gene trap lines, the continuing evolution of gene trap vectors, and its easy accessibility to scientific communities, gene trapping could provide a fast and efficient way of generating mouse mutation(s) for any one particular gene of interest or multiple genes involved in a pathway at the same time. Consequently, we recommend gene trapping to be considered in the planning of mouse modeling of human reproductive disease and the IGTC be the first stop for people interested in searching for and generating mouse mutations of genes of interest.

Gene trapping is a high-throughput approach of generating mutations in murine ES cells through vectors that simultaneously disrupt and report the expression of the endogenous gene at the point of insertion. First-generation vectors trapped genes that were actively transcribed in undifferentiated ES cells. Depending on the areas in which they integrate, these vectors can be roughly divided into two classes:

  • promoter trap vectors and
  • gene trap vectors.

Promoter trap vectors contain promoterless reporter regions, usually bgeo (a fusion of neomycin phosphotransferase and b-galactosidase), and thus have to be integrated into an exon of a transcriptionally active locus in order for the cell to be selected for neomycin resistance or by LacZ staining. Gene trap vectors demonstrate more utility by their added ability to integrate into an intron. These vectors contain a splice acceptor (SA) site positioned at the 50-end of the reporter gene, allowing the vector to be spliced to the endogenous gene to form a fusion transcript. Later improvements include an internal ribosomal re-entry site (IRES) between the SA site and the reporter gene sequence; as a result, the reporter gene can be translated even when it is not fused to the trapped gene. Second-generation vectors have sought to trap genes that are transcriptionally silent in ES cells. Although these vectors still contain a promoterless reporter gene with a 50 SA sequence, the antibiotic resistance gene is under the control of a constitutive promoter. Consequently, antibiotic selection is independent from the expression of the trapped gene, whereas the expression of the reporter gene is still regulated by the endogenous promoter.

A disadvantage of these vectors is that all integration events give rise to resistant ES cells regardless of whether or not the vector has integrated into a gene locus. To increase trapping efficiency, a new class of polyA gene trap vectors was developed where the polyadenylation signal of the neo gene was replaced by a splice donor sequence, thereby requiring the vector to trap an endogenous polyA signal for expression of neo. These vectors were recently shown to have a bias toward insertion near the 30-end of a gene due to nonsense-mediated mRNA decay of the fusion transcript. An improved polyA trap vector, UPATrap, was developed to overcome this bias using an IRES sequence placed downstream of a marker containing a termination codon. Gene trap vectors are usually introduced by retroviral infection or electroporation of plasmid DNA, with each approach having its own advantages and disadvantages.

While relatively difficult to manipulate, retroviral gene traps display a preference toward insertion at the 50-end of genes, which is advantageous for generating null alleles. Moreover, the multiplicity of infection with retroviruses can be tightly controlled to a single trap event or simultaneous disruption in many genes. However, there may be a possible bias integration toward certain ‘hotspots’ of the genome.

In contrast, plasmid-based gene trap vectors integrate more randomly into the genome. This can, however, potentially result in a functional partial protein and a hypomorphic phenotype. Additionally, plasmid vectors usually result in multiple integrations in 20–50% of cell lines. The most common approach for identifying the gene trap integration site is to use 50 or 30 rapid amplification of cDNA ends (RACE) to amplify the fusion transcript. The sequence provides a DNA tag for the identification of the disrupted gene and can be used for genotypic screens. Mutagenesis screens can also be performed on the basis of gene function or expression, and data from an expression sequence combined with sequence tag information can elucidate novel expression patterns of known genes or to suggest gene function.

Gene trapping has proven to be an efficacious technique in mutagenesis compared with other methods such as

  • spontaneous mutations,
  • fortuitous transgene integration and
  • N-ethyl-N-nitrosurea (ENU) mutagenesis

We have been able to use our SpiderGene program to identify genes in reproductive tissues that are present in the IGTC database and moreover to narrow down those with restricted expression in the testis and ovary. Gene trapping possesses an enormous potential for researchers in the reproductive field seeking to create mouse models for a gene mutation. The improving versatility of gene trap vectors has enabled groups to trap an increasing number of genes in various organisms, including Arabidopsis, Zebra fish and Drosophila.

The gene trap effort has perhaps been the most extensive in the murine genome, with over 57000 cell lines representing more than 40% of the known genome. These large-scale screens will likely achieve the trapping of the entire mouse genome in the coming years, but the power of gene trapping will only be fully demonstrated by its usefulness in investigator-driven focused functional analyses.

In our laboratory, future work will focus on generating knockout mice in order to investigate gene function and to identify gene products that might have therapeutic value in reproduction. As screening efforts continue, gene trapping will continue to be a valuable tool in mouse genomics and will undoubtedly yield new discoveries in Reproductive Physiology and Pathology.

Source References:


Read Full Post »

Author and Reporter: Ritu Saxena, Ph.D.

On June 4, 2012, I authored a post on HBV and HCV-associated Liver Cancer: Important Insights from the Genome reporting about the major role of chromatin remodeling complexes and involvement of both interferon and oxidative stress pathways in hepatocellular malignant proliferation and transformation based on the genes showing recurrent mutations in the observed genes.

In this post, I have discussed the latest research on cyclin B1 and Sec62 expression in PBMCs of HCC patients and how their elevated expression correlates to significantly to negative prognostic value in terms of recurrence-free survival.

Researchers at the Changhai and Gongli Hospital in Shanghai, Military Medical University, People’s Republic of China recently identified the candidate biomarkers for HBV-related HCC recurrence after surgery. The research was published in the June 2012 issue of Molecular Cancer journal. According to the group findings, Cyclin B1 and Sec62 may serve as effective biomarkers and potential therapeutic targets for HBV-related HCC recurrence after surgery.

Research article: Identification of cyclin B1 and Sec62 as biomarkers for recurrence in patients with HBV-related hepatocellular carcinoma after surgical resection.

HCC background and Research Problem: Hepatocellular carcinoma is cancer of the liver. It is different from Metastaticc liver cancer, which starts in another organ (such as the breast or colon) and spreads to the liver. The most frequent factors causing HCC include chronic viral hepatitis (types B and C), alcohol intake and afla- toxin exposure.

In most cases, scarring of the liver referred to as cirrhosis is an important risk factor for HCC. Cirrhosis may be caused by:

However, patients with hepatitis B or C are at risk for liver cancer, even if they have not developed cirrhosis.

According to the data from International Agency for Research on Cancer, hepatocellular carcinoma (HCC) is the third leading cause of cancer-related deaths worldwide, with over a half million deaths per annum.

In China, a very high infection rates with HBV have been reported. According to the recent statistics reported by Jemal et al in 2011, HCC cases occurring in China account for 55% of the total cases reported in the world.

Surgical resection, although provides an opportunity for cure, however, frequent recurrence post surgery has posed a major challenge to longterm survival. Pertinent to their research, authors state “Frequent tumor recurrence after surgery is related to its poor prognosis. Although gene expression signatures have been associated with outcome, the molecular basis of HCC recurrence is not fully understood..”.

Research: To determine the molecular basis of HCC, authors used the Peripheral blood mononuclear cells (PBMCs) to predict the recurrence of HCC after surgery. Use of PBMCs was in contrast to previous studies that used just the liver tissues. PBMCs have the advantage of being easily obtained in the clinical setting. Thus, identification of biomarkers using PBMCs would be a great way to predict the recurrence of HCC post surgery.

A microarray-based gene expression profiling was performed to indentify candidate genes related to HCC recurrence. In all, mRNA derived from 6 HCC cases (3 cases with recurrence and 3 without recurrence) were subjected to genome-wide analysis. Some critical genes were indentified including cyclin B1 (CCNB1), SEC62 homolog (S. cerevisiae) (SEC62), and baculoviral IAP repeat-containing 3 (BIRC3), suggesting that they probably play important roles in the pathogenesis of HCC recurrence. To confirm the results of microarray analysis, the mRNA and protein expressions of these 3 genes were measured in 80 HCC samples from HCC cases and 30 samples from healthy cases. The authors found that the transcriptional and protein expressions of cyclin B1, Sec62, and Birc3 in the PBMCs were significantly higher in HCC samples than those in the non-recurrent and normal samples.

Furthermore, to determine the clinicopathologic significance of cyclin B1, Sec62, and Birc3 in HCC, immunohistochemical analysis from 35 recurrent tissues and 45 nonrecurrent revealed that the protein levels of cyclin B1, Sec62, and Birc3 were substantially higher in the recurrent tissues than those in the non-recurrent samples. Thus, the immunohistochemical results from tissues were consistent with the previous transcriptional and protein results in PBMCs.

Conclusion of study:  The authors discussed that “In recent years, studies on malignant tumors has primarily focused on cell proliferation, migration, and apoptosis. Cyclin B1, Sec62, and Birc3, chosen in this study according to our microarray analysis, likely play important roles in cell proliferation and migration. They can exert a tumor-promoting effect on HCC by regulating cell cycle and protein translocation.” As derived from the statistical methods employed in the research, elevated cyclin B1 and Sec62 expression in PBMCs had a significantly negative prognostic value in terms of recurrence-free survival, which hints the potential use of these molecular markers to predict the risk of tumor recurrence after surgery and to act as therapeutic targets to reduce tumor recurrence and improve clinical therapies.

Thus, these results revealed that cyclin B1 and Sec62 may be candidate biomarkers and potential therapeutic targets for HBV-related HCC recurrence after surgery.

Read Full Post »