• Connected Data Ecosystem – Project is FAIR
• Data shareable
• NIH – agenda on data: diverse sets of data: Images of MRI, cells, of organs, of communities,
• Share images and link it to tables
• METADATA 34PB enable search – moving Data to clouds for Large-Scalable Analysis
• Sequence Read Archive (SRA) – DNA seq.
• COVID-19 from around the World SRA in Cloud Partnerships enabled
• Open Science – enhance SW tools for making research cloud-ready
• NIH has 12 Centers: Genomics, Neuro-imaging
• SCH – Smart & Connected Health
• IT, Sensor system hardware, effective usability, medical interpretation, Transformative data Science
• Cancer, Alzheimer’s, Genomics, Medical Imaging, Brain circuits,
• Coding it Forward: Students come to NIH Virtually from home to join CIVIL DIGITAL FELLOWSHIP
• COVID-19: repositories of data for researches:

1. Treatment for Interventions
2. Long term Sequelae
3. Clinical platforms: BigData Catalyst, Allow US, ADSO, National COVID Cohort
4. Across platforms: workflow after RAS August Deploy: Passport for researchers to access data faster, Privacy-Preserving Tokens, Interoperability across clinical COVID data bases
5. Metadata super rich to link to other new data sources is a challenging issue to solve across studies

Scott Parker

Sinequa Corp

Director of Product Marketing

• Disconnect between R&D & IT
• Intelligence search Applications for sensitive information: Sinequa is a leader
• shares one index cost for document go down & productivity increases

Ari Berman

BioTeam Inc

Chief Executive Officer

• Distributed Questions from the Audience to the speakers

10:00 AM – 11:25 AM EDT on Tuesday, October 6

Etzard Stolte

Roche Pharma

Global Head

October 7, 2020

The Chicagoland COVID-19 Commons: A Regional Data Commons Powering Research to Support Public Health Efforts

• 

  Matthew Trunnell

  VP & Chief Data Officer

9:00 AM – 9:20 AM EDT on Wednesday, October 7

• Seattle & COVID – samples from Seattle Flu Study
• Public Health Practice vs Research – Data from Human Subjects: Avoid delute the control
• Chicagoland COVID-19 Data Commons – in Chicago

1. Neighborhood level in Chicago
2. common data model
3. power efforts Predictive modeling : Case rate Total confirmed cases, Death cases
4. Legal agreement of the Consortium
5. https://chicagoland.pandemic

• Commons – resources held in commons non-for profit
• Data Commons: cloud based SW platforms that are co-located data, computing infrastructure and applications
• Level 1: Basic, Level 2: Repeatable, Level 3: Governance Level4: Interoperability Level 5: Sustainable
• COVID-19 Data Common: Public health authorities collects data – nor available to Research community
• Research community need access to Public health authorities
• Regional COVID-19 Data Commons: Reasons: Public health decision is LOCAL but specific to the Region
• Fund raising in the communities
• Data 1: Clinical Data for Health care Summary of incidence – Signals of ethnic dependencies and co-morbidities

1. Safe harbor: removal of 18 identifiers
2. Expert Determination

• Data 2: Public Data: Environmental,
• Data 3: Resident-Reported Data on iPhones: multiple languages supported early reports of people feeling unwell

CompBio: An Augmented Intelligence System for Comprehensive Interpretation of Biological Data

Richard Head

Washington Univ

Prof & Dir Genome Technology Access Ctr

9:20 AM – 9:40 AM EDT on Wednesday, October 7

• Formating, data scrubbing,
• Replace data fabric with simplified version
• create “Memory Model” Machine learning does classification of patterns
• dimensions are the variables
• “Hyper-dimensional – ingestions of abstracts and articles
• Example; IL^: Aggregate Memories to create a NORMALIZED Aggregate Memory
• Relationships explored
• Complex Knowledge Patterns Generated by the PCMM: Compared Utilization
• Augmented AI System: Combination PCMM with AI
• Literature mining CompBio
• Evidence of Utility: PCMM – Accepted or Published Research Leveraging PCMM Applications
• Example 1: Cell Metabolism CompBio – A person formulate hypothesis
• Example 2: Analysis of RNA-Seq a rare mutational subtype of GBM

1. Hypothesis –>> BioExplorer –>> Multiple relations revealed
2. Example 3: Animal Models to Human Disease: CompBio – Crohn’s Assertion Engine

Summary – Augmented AI Platform for Biological DIscovery

• PCMM – Memory modle – hyperdimensional
• AAI Infrastructure
• Knowledge map libraries
• In development Medical Discoveries

• Cancer Classification: Shift from Anatomy/History of Molecular Etiology
• Chronic Myeloenous Leukemia  – Gleevec
• Type cancer seq:

1. Hereditary cancer sequencing – BRCA
2. Tumor cancer sequencing

• Panel Sizes – 500-1000x – the bigger the panel – more computational time more data need be investigated

1. Hotspot Panels,
2. Gene Panels,
3. Exomes

• Cell free DNA Testing – Liquid biopsy

1. Apoptosis
2. Necrosis

• FoundationONE
• Patient Results: ALL mutations found, Mutation Burden,
• Gene EGFR – no mutation
• For every Mutation what Therapy is recommended for approved drugs
• Clinical Trials for the mutations
• VARIANTS of unknown significance
• WORKFLOW: many MDs send sample get 38pps report
• Genomic Classification and Prognosis in AML: Mutations subset and therapies available
• Paradigm Shift in Classification

1. 2013 – Lung Adenocarcinoma <<<- –
2. 2011 – another cancer

 

mTOR System: A Database for Systems-Level Biomarker Discovery in Cancer

• 

  Iman Tavassoly – CANCELLED

  C2i Genomics

  Physician Scientist

10:20 AM – 10:40 AM EDT on Wednesday, October 7
Add to Calendar

mTOR system is a database I have designed for exploring biomarkers and systems-level data related to mTOR pathway in cancer. This database consists of different layers of molecular markers and quantitative parameters assigned to them through a current mathematical model. This database is an example of merging systems-level data with mathematical models for precision oncology.

FAIR and the (Tr)end of Data Lakes

• 

  Kees Van Bochove

  The Hyve

  Founder & Owner

10:20 AM – 10:40 AM EDT on Wednesday, October 7

Normalizing Regulatory Data Using Natural Language Processing (NLP)

• 

  Qais Hatim, Dr.

  FDA CDER

  Visiting Assoc

David Milward

Linguamatics

Senior Director, NLP Technology

10:40 AM – 11:00 AM EDT on Wednesday, October 7

• ML focus on Disease
• NLP – different words have same meanings, different expression same meaning, grammer & Meaning
• Normalizes output

1. Disease
2. Genes
3. Dates
4. Mutations

• Transform Unstructured into structured
• Identifying Gaps in adverse events Labelling: Pain and Opioids
• Improve drug safety
• ChemAxon

Supplemental Approval Letters

Coding for Adverse events: “derived values of possible interest”

• Use of Prominent Terminologies used at the FDA: UNII – Translation into ANSI tesaurus standard
• Matching to the Variation found within Real Text: synonyms
• Using ML for Normalization in Disease Context
• Deep Learning PRE-TRAINING APPROACH for annotated date = supervised learning
• A set of rules to handle overlapping entities
• normalized the amp extracted from concepts
• BERN and Terminologies: BioBERN, PubMed Central, PubMed Articles
• NER – Named Entity Recognition
• Evaluation of the Approach

Conclusions

NLP, ML, Hybrid methods, Terminology +ML methods

Building an Artificial Intelligence-Based Vaccine Discovery System: Applications in Infectious Diseases & Personalized Neoantigen-Related Immunotherapy for Treatment of Cancers

• 

  Kamal Rawal

  Amity Univ

  Assoc Prof

10:40 AM – 11:00 AM EDT on Wednesday, October 7

• Classification of proteins
• Data Collection
• Feature Selection – Most important from 1447 features
• Deep learning Model: Vaxi-DL: Layers, compilation
• Overfitting Model strategy
• Balancing Imbalanced
• Hyper parameter tuning: Internal parameter of the model
• Stratified K-Fold Training and Validation
• Ensembling Approach: many weak classifier to create a STRONG Classifier
• ROC Curve: Ensemble by Consensus
• Before and after calibration
• Benchmarking the system: Vaxi-DL Ensemble by Average vs by Consensus
• SYSTEM developed: Type protein – find results
• Rare disease CHARGE Syndrome was used for validation
• Application to COVID-19 – Methodology
• Application on Cancer: Which peptide can be used as antigen for prediction of immunogenic peptides

 

Using GPU Computing to Evaluate Variant Calling Strategies

• 

  George Vacek

  NVIDIA Corp

  Sequencing Strategic Development

• 

  Eriks Sasha Paegle

  Dell EMC

  Senior Business Development Manager

11:15 AM – 11:30 AM EDT on Wednesday, October 7

• Navidia: 100 Genomes Cohort generated at NY Genome Center  NHGRI
• Navidia Parabricks mentioned AZURE
• Dell EMC: Test environment: Dell Technology Cloud Storage for Multi-Cloud: resources across GCU, AWS, Azure in Northern Virginia regions
• Multi-Cloud ease of use: without Multi-cloud vs with Faction multi-clouds
• Ease of use
• Deep Averaging Network (DAN)
• NVIDIA CLARA PARABRICK TOOLKIT: Short & Long read, Deep learning, Data Analytics, ML
• Reference applications – host of customized applications, 3rd Party App, Libraries
• GPU (Genomics PUs) – Drop in tools for Somatic Pipelines : Clara Parabricks v3.5
• Partnership of NVIDIA and Petagene announced at BioIT20 – NGS Data compretion
• Petagene technology allows lossless compression reduce storage costs
• Project with Sanger Institute – Optimizing Muto-graph Identification
• completed run in 24 hours instead 31 days
• Parabricks is a joint project Dell/EMC and NVIDIA

PLENARY KEYNOTE: Game On: How AI, Citizen Science, and Human Computation Are Facilitating the Next Leap Forward

12:30 PM – 1:55 PM EDT on Wednesday, October 7

• 

  Allison Proffitt

  BioIT World & Diagnostics World

  Editorial Dir

Seth Cooper

Northeastern Univ

Asst Prof

• Foldit – Scientific discovery using video games in the domain of protein structures and folding
• Combine Human with machine
• Score based on competition among players for higher score and collaboration in groups
• Problem: Chemistry give input.
• Puzzle available for one week on the Internet, games ongoing,
• Solution analysis – continually IMPROVE the structure of Protein folding
• Foldit Tutorials offered online
• Player accomplishments: Articles by scientists ,
• development of algorithms discovery
• Electron Density fitting
• Enzyme re-design
• de novo Protein Design – named authors on a paper – scientific process
• Future Work: Coronovirus Spike protein
• Small molecule design
• narrative
• virtual reality – 3D protein structure for manipulation
• htp://Fold.it/Educator Mode
• htp://Fold.it/standalone
• http://fold.it/
• seth.cooper@gmail.com

Lee Lancashire, CIO

Cohen Veterans Bioscience – not for profit – advancing Brain health

• Biotyping and stratification
• Biomarkers
• Omics data
• All meet in the Common – Brain Commons: Clinician, Geneticist, Scientist, Bioinformatician, R Studio, Python, Jupyterhub
• Multidimensional Biomarkers in Multiple Sclerosis

 

Pietro Michelucci

Human Computation Institute

Director

• Why machine can’t tackle AI on their own and AI can’t do Precision Medicine on their own
• young people more than others N of 1 – Precision Mediicne
• Scandinavians and Russians are immune
• AI & Precision Medicine: can’t solve the complexity of messy data vs big data
• Messy data: heterogeneous multidimensional, to many combinations to explore, select which combination to explore vs let the machine generate all the combination and do analysis on all and discover PATTERN
• Causal vs spurious
• Logical reasoning, right brain abstract and short cuts – Human brain does routinely
• Human do better on context: Not all info is in pixels such as context
• #ADS – SBIR suspected the hypothesis to be tested
• improving crowd wisdom methods: 20 input by different people PLUS machine
• combine crowd answers with machine faster and improved accuracy
• Machine has no intuition – machine bias of Human and of machine is similar
• Wisdom of Crowd: Bootstrapping hybrid Intelligence: CIVIUM
• bit.ly/civiumintro

 

Kevin Davies, PhD

CRISPR Journal

Exec VP & Exec Editor

Timothy Cutts

Wellcome Sanger Institute

Head

Vivien Bonazzi

Deloitte Consulting LLP

Managing Dir & Chief Biomedical Data Scientist

• How organizations use bioscience data
• Data Ecosystem: Hardware and software: Cloud and other options
• Operationalize the two trends:

1. Platforms: End to end solutions resulting in SILOS, systems are native: data ingestions
2. Data Commons: Open arch, open source – integration and interdependence issues

• Biomedical Agencies in NIH various Organizations in the Private sector: Sharing data must be more effective
• IT, Data Science, Management – COVID – reduced barriers
• Leadership: Different voices from different people
• Data strategies & Governance not the whole but small pieces , incentives to share data

Chris Dagdigian

BioTeam Inc

Sr Dir

• 10th Anniversary to Trends from the Trenches
• IT infrastructure changes
• Research IT:

1. Genomics & BioInformatics
2. Image-based data acquisition and analysis: CryoEM, 3D microscopy, fMRI image analysis
3. ML and AI – GPU FPGAs, neural processors: Drive in organizations: bottom up
4. Chemistry & Molecular Dynamics
5. Storage and exploitation of data for insights
6. 2020 Hype vs Reality
7. Scientific Data: managing and understanding, data movement, federated/access
8. Big Data: data storage, management & governance standards vs human curated data
9. IT needs guidance and decisions from Science Team
10. Culture change for joint management by Science & IT: data fidelity, attribution, allocation top down
11. NERSC File System quotas & Purging overviewSilos & So
12. Petabytes of open access data, collaborative research resources: Data rich environments
13. Data Lakes: Gen3 Data Commons
14. Data hygiene:metadata is Science side vs IT
15. Biased Data: Model & Data Bias

• Failed Predictions:

1. Compilers matter again – not True
2. CPU benchmarking is back – WRONG
3. AMD vs Inter arm64 vs both
4. Policy driven auto-tiering storage – wrong, USER self-service for tiering, movement and archive decision. Let researchers tier/move/archive based on Project, Experiment or Group
5. Single storage namespace – Wrong: Data intensive science: scientists must do some IT jobs themselves

Kjiersten Fagnan

Lawrence Berkeley Natl Lab

CIO

• Genome Project of DOE
• Data management with other agencies
• COVID: Collaborations, breaking down barriers, small labs and big labs ALL generate data and sharing
• that collaboration is needed regardless of COVID – not happen
• If twoo big one lab can’t handle it all
• Funding and training does not support the Collaborations because next round of funding depend on individual publications – which requires silos
• Data cleaning and data management:Standards are annoying and painful – not needed for publishing the results as soon as possible – just that someone else will be able to use it
• Facebook have hundred of curators – the curation of scientific data requires same hunsrands od curators that are SCIENTISTS and Data scientists

Matthew Trunnell

Pandemic Response Commons, Seattle

VP & Chief Data Officer

• Data commons for intra- and inter-mural data sharing
• ML is needed for Data commons
• Progress in FAIRness, NIH efforts driven by Susan Gregory across NIH all centers
• Large amount of B-to-B Data sharing UBER sharing with a jurisdiction they operate
• SNOWFLAKES – new cloud technology
• COVID – plays an accelerator
• Cancer vs COVID – transfer knowledge from COVID to Cancer

9:00 AM – 10:40 AM EDT on Thursday, October 8

11:35 AM – 12:00 PM EDT on Thursday, October 8

Robert Green

Brigham & Womens Hospital

Co-founder of Genome Medicine

Prof & Dir G2P Research

• Combining data to rapidly analyze COVID-19 Patients –
• identify BIOMARKERS for vulnerability
• Preventive Genomics – Angelina Jolly’s musectomy as a preventive clinical condition
• Patients access to own genomics data
• Population screening – to predict risks
• Genetic Testing to Consumer: Preventive Genomics: conflated genotyping/sequencing and labs/care providers
• Genetic Testing to Consumer: COST & Benefits – UNCLEAR

1. diagnosis of unsuspected genetic disease
2. stratification for surveillance
3. which pieces of the puzzle need to be brought to bear in patient care
4. Categories and Reporting criteria: Gene-Disease validity vs Variant Pathogenicity –>> Clinic
5. MedSeq Project: 10MM randomized study – all genome info shared with Patient, other arm only selective genome data shared with patient: 100 patients 20% carry monogenic condition: Polygenic risk scores:
6. CAD – high Cholesterol biomarker, A-FIb, DM2, 52% Women 48% Men
7. No high risk error by PCP discussing and disclosing the results of the sequence
8. Filtering the results: Indication -based testing vs Screening
9. BabySeq Project: INFANTS sequencing to prevent disease: 11% carry a mutation in a monogenic gene for a monogenic condition -like abnormal narrowed aorta
10. MDR – Monogenic Disease Risk
11. MilSeq Project: US Air Force – Military active duty
12. 5,8,10 – are all Polygenic studies
13. Polygenic Risk Scores – High risk
14. Classification need to be repeated every few years (2 years – re-sequence) due to changes in health and to efficiencies in new discovery in curated data which is improving as on-going

• Risk benefit – UTILITY – Partners Biobank Return of Genomic Results
• No interest on knowing by the Public NCCN criteria on chart review 20%
• Brigham Preventive Genomics via telemedicine – First in the country
• APC mutation after colonoscopy – obstruction diagnosed
  www.genomes2people.org 
• @robertgreen

Juergen Klenk

Deloitte Consulting LLP

Principal

• Bradykinin hypothesis for COVID-19
• liberate the data: People , Data Risk

Natalija Jovanovic

Sanofi

Chief Digital Officer

• AI in Pharma
• Vaccine preventable diseases – produce 1Billion vaccines a year

1. reduction of incidence: Pertusis – 92% eradication

• manage risk profile
• Science mechanism translatable to machines

1. high automated ingestible data for AI
2. Digital is about people: Good data Good algorithms Good GUI

Vivien Bonazzi

Deloitte Consulting LLP

Managing Dir & Chief Biomedical Data Scientist

12:00 PM – 1:25 PM EDT on Thursday, October 8
Add to Calendar