Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
U of Minnesota Psychiatry/Psychology (Alik Widge, MD, PhD)
Teaming Overview and Objectives
Project Overview
• Deep bench of expertise in cognitive profiling, including assays proven as test-retest stable
• Access to large cohorts (1000s) of patients receiving brain-based interventions to optimize mental function
• Supercomputing facilities, deep learning/informatics expertise, on-campus medical device prototyping
• Success in past DARPA programs (SUBNETS, ElectRx) requiring similar phenotyping, profiling
• Need partners in: genomics, metabolomics, wearables, other non-brain quantification
• Core concept: success in specialized roles is driven by mental performance/readiness, not physical
• Design a Reliable, Repeatable, and Robust framework for tracking mental performance capacity
• Multi-domain: self-control, motivation/drive, working memory, arousal regulation, etc.
• Link mental performance to physiologic/expressomicmarkers that can be tracked in real time
Impact• Wearable or minimally implantable personal ”brain
performance monitor”, totally passive and transparent• Measure individual cognitive capability (mission readiness)
on a day-to-day, hour-to-hour timescale• Optimize not just execution, but training, use of recovery
time, overall psychological health/well-being• Envision a consumer/business-facing performance
optimization technology as a major output• Identify high-potential leadership candidates• Optimize coaching of inconsistent performers• Maximize educational/training outcomes
Peripheral Markers
Optimal Performance
Self-Control
Drive to Win
Focused Mind
CircuitsRhythms Skills
Brain Health and Capacity
Alik Widge, MD, PhD | [email protected] | 612-625-7594
David Putrino, Icahn School of Medicine at Mount Sinai, Abilities Research Center
Teaming Overview and Objectives
Project Overview
• Multiple partners inside and outside of the institution• >50 peer-reviewed publications, book chapters and
presentations. • International reputation for training some of the best
athletes in the world• Core facilities (in addition to everything medical):
• Motion capture• Genomics and microbiome• AI and data science core facility • Tech transfer capabilities
• Looking for collaborators who can assist with device fabrication and development, rapid prototyping, bioengineering
• Access to elite athletes, high performance executives and special forces soldiers
• We have a strong background in tracking movement and behavior in both human and non-human models
• Our goal is to quantify less traditional predictors of performance and monitor how they vary during different high performance conditions
• We have been investigating social, emotional and cognitive predictors of human performance in multiple settings
Impact• Our primary outcome measure will be to show that
performance prediction is enhanced by adding objective social, emotional and cognitive to our predictive model
• Enhanced ability to identify and organize performers into effective teams
• Quantify the effect of different social emotional features on daily performance
• If successful, translate findings into a platform that can be replicated
[email protected] – 617 955 4364
Louisiana State University System and CollaboratorsPennington Biomedical Research Center LSU Athletics LSU School of Kinesiology
Teaming Overview and Objectives
Project Overview
Existing team members and partners Pennington Biomedical Research Center LSU Health Sciences CenterLSU Athletics Baton Rouge General Sports Med.LSU School of Kinesiology Baton Rouge OrthopedicsNASA, USARIEM, WRAIR Our Lady of the Lake Neuroscience
Relevant experience• Highly-trained, proven research and athletics training teams with extensive military,
external, and private collaborators• $80 million in DOD funding over 30 years with > 100 peer-reviewed publications• Single and multi-center NIH clinical trial experience including MoTrPAC, LIFE, et. al.• >15 past and ongoing funded research projects that utilized elite athletes
Institutional assets• Internationally-recognized genomics, proteomics, metabolomics, immuno-metabolomics,
microbiome, bioinformatics, behavioral/cognitive, whole-body phenotyping, and bioinformatics cores
• Performance phenotyping and intervention laboratories for exercise testing and training• World-renowned recruiting and marketing core• Interventional resource management core with stellar participant retention/compliance
For which technical challenges are you seeking collaborators?• Access to “machine-learning” level data processing to comprehend expression circuits
Does your team have unique sample/cohort access?• Access to elite professional, collegiate, and high school athletes
• Elite athletes posses the genetic and phenotypic characteristics, including cognitive/behavioral aptitude, physical prowess, and resiliency, similar to special forces personnel.
• Our research team has access to elite athletes, and the research experience and facilities that will allow us to examine the genetic, physical, cognitive, and behavioral elements (expression circuits) present in high school or early collegiate athletes, implement specialized training and nutritional interventions, and validate developed prediction models for successful attainment of “professional“ status.
• For all funded proposals in TA3, the major technical challenge will be to recruit, test, and intervene on potential special forces quality individuals to validate products from TA1 and TA2.
ImpactAlter the way the special forces select and train soldiers to enhance the health and performance of warfighters.Potential applications enabled by this technology:
• Identify individuals who, if desired, can fast-track in the military to special forces.• Select treatments to promote health and performance, and select individuals to foster team
adhesion, in collegiate and professional athletics.• Offsite diagnosis and monitoring of patients with health-related concerns or who are at risk
of future disease. • Allow individuals in the private sector to track their own health related outcomes.
Unique metrics and milestones:• Multi-center collaborations with TA1/TA2 sites• Continue fruitful collaborations with analysis cites for metabolomics• Excellent participant recruitment and protocol compliance• Near real-time electronic data entry and QA
How will the team pursue transition of this technology?• We will effectively collaborate with the TA1 and TA2 teams to implement the technology in
participants including high level athletes and participants from the greater BR area. • Technologies validated in TA3 will be cross-validated in the DOD through our existing
collaborations with NASA, USARIEM, WRAIR, and others.
Jennifer Rood [email protected] Jack Marucci [email protected] Neil Johannsen [email protected]
PI: J. Craig Venter, Venter Institute Center for Human PerformanceTeam: UCSD, Caltech, Metabolon, NSI, Beyond Limits
Team: J. Craig Venter (PI), Pamila Brar (Chief Medical Officer), Maria Giovanni, Ewen Kirkness, Todd Michael, Marcelo Freire; AI: Gary Fogel, NSI; AJ Abdallat, Mark James, Beyond Limits; MRI and Neuroscience: Jim Brewer, Anders Dale, UCSD; Implantable Devices: Axel Scherer, Azita Emami, Caltech.Experience: 25 years pioneering human genomics, sequenced first free-living organism, first human genome, first diploid genome, first microbiome, and created first synthetic cell; >10,000 human genomes analyzed; performed first integrative study of microbiome / metabolome / human genome with predictive analysis.Assets: Venter Institute Center for Human Performance - a genotype-phenotype platform to comprehensively access an individual’s physical, cognitive, physiological and behavioral traits and accurately sequence their genome; using whole body and brain MRI; behavioral, cognitive and physiological tests; metabolome and microbiome assays; large scale predictive analytics capabilities.
J. Craig Venter - [email protected] - 858-200-1890
Large scale whole genome sequencing
10,000 human genomesTelenti et al., PNAS 2016
Genotype-phenotype integration, predictive AI
analytics
Facial predictionLippert et al., PNAS 2017
Whole body & brain MRIHou et al., BioRxiv 2018
Advanced imaging technology
The integrated
platform will lead to direct
phenotype predictions
from the genetic code
Human microbiome predictive health
Metagenome of fatty liver diseaseLoomba et al., Cell Metab 2017
Success, Technology, Milestones, Transition:• Sequenced and de novo assembled diploid genomes to produce complete and accurate
genome data.• Extensive phenotype characterization from imaging to cognitive testing for
comprehensive and complete human representation.• Artificial intelligence (AI) including deep learning approaches to enable prediction of
phenotypes from the genome sequence.• Not-for-profit and for-profit entities to provide best in class genotype-phenotype
services; computational tools and models for prediction of human performance and disease; nano-sensing body fluids with implantable technologies for elite performers.
1 billion data points
Mohammed Eslami, Netrias
Teaming Overview and Objectives ImpactTA1 omics learning & data integration (Netrias), cross-validated
nested phenotypic assays (Nascent), bioinformatics (NYU),active & adaptive learning (PARC)
TA2 wearable biometric and continuous saliva sensing;low-power wireless electronics; sensor fusion & analytics;UX/UI design; system integration (PARC); real-time in-vivosensing (UCSB)
Relevant Experience (Netrias/Nascent/NYU)Transdisciplinary assessment for Special Forces & astronautsMachine learning and Data integration of ‘omics dataLeaders in the DARPA SD2 Program/NIH CPTAC/NASA HRP
Assets (Netrias)Active Discovery EngineTM for rapid data integration/curation
Desired partnersAdditional experience with in-vivo molecular sensing
Unique sample/cohort access: No
Deep biological structure of phenotypic assays revealed by management and integration of heterogeneous measurements from multiple layers and time scales that are concurrent, interrelated, and non-interfering.
TA1/
2TA
2
You?
TA1
Nascent S&T
Deep Phenotyping Rapid Integration and CurationAI/ML
ML-Ready Data
Physio-logical
Behavioral Cognitive
Social
Anticipated Impact Continuous cross-validation of nested phenotype measures for rapid tuning to individual differences and situations (adaptive personalized assessment); Automated data integration & curation
Unique metrics & milestones 10x reduction of uncertainty in personalized assessment for uniquely objective selection, >95% accuracy of expression circuit → phenotype prediction, 80% reduced time for data integration
ApplicationsPrecision health; medical home model; predictive diagnostics and assessment for personalized training; Train the Trainer
Technology Transition Netrias works with pharmaceutical companies to transition infrastructure & analytics to develop their drug and s/w offerings
Nested scalesof assays
Mohammed Eslami – [email protected] – 202.213.0191
TA1 Netrias: omics learning & data integration Nascent S&T: phenotypic assays NYU: bioinformaticsPARC: active & adaptive learning
TA2 PARC: wearable biometric, continuous saliva, wirelessly powered sensing low-power wireless electronics sensor fusion & analytics UX/UI design, system integration
UCSB: real-time in-vivo sensing
Technical Areas & Teaming
Project Overview
Relevant Experience (PARC)Development & deployment of wearable sensors, Data analysis from dispersed sensors & Remote RF power,Success in DARPA Sensor Tape & Explainable AI, ARPA-E MONITOR programs
Assets (PARC)Cleanroom/fabrication facilities for sensor/device manufacturingComputing assets for AI/ML
Desired partnersAdditional experience with in-vivo molecular sensingMedical device manufacturers with regulatory experience
Unique sample/cohort access: No
Anticipated impact Real-time sensing: warfighter situational & performance awarenessActive & adaptive learning: customized training programs
Unique metrics & milestones Real-time molecular detection in-vivoUnobtrusive biometric sensingBiomarker selectivity and sensitivity
ApplicationsFirst-responders, high-performing athletes & healthcare
Technology TransitionPARC’s business model focuses on commercialization of new technologies; our in-house business development team works with a wide range of commercial interests across fields
Advanced Sensing
AI/ML
Communication
UI/UX
Palo Alto Research Center – Hoda Eldardiry, Sean Doris
TA1/
2TA
2
You?
TA1
Nascent S&T
Objective PARC proposes a real-time molecular target monitoring platform to track biomarkers identified by TA1
Hoda Eldardiry, Ph.D. – (650) 812-4790 – [email protected]; Sean Doris, Ph.D. – (650) 812-4242 – [email protected]
Machine Learning
Rich Interactive Platform
Knowledge Base
Expert User
Input
Teaming Overview and Objectives
Project Overview
• Vincent Emanuele, PI○ Former visiting scientist at CDC for 6.5 years, head of data science R&D
at Wellcentive○ Experience in interdisciplinary machine learning R&D (proteomics,
genomics, physiological sensors, patient medical records)• Yiftach Eisenberg
○ Former DARPA PM/Deputy OD and AI thought leader within DARPA○ Experience in R&D related to machine learning problems in national
security• James Cheng
○ Senior machine learning researcher, former Georgia Tech Postdoc, and researcher at Wellcentive
○ Has successfully researched and commercialized ensemble learning systems on a large scale dataset of 30M patients
• Bryant Menn○ Biomedical engineer (Georgia Tech)○ 3+ years experience working as both a data engineer and data scientist
to support machine learning research
• We’d like to combine our extensive experience in machine learning research with collaborators who are strong on the lab side for TA1 and TA2
• We have experience in proteomics, genomics, physiological sensors, patient medical records, and mobile sensing
• TA1: data analysis and new algorithms surrounding phenotypic assays / physiological output
• TA2: Data visualization platforms for real-time molecular target monitoring
Impact• We go from white board to software prototype and work
on large datasets• We deliver machine learning prototype solutions that
integrate well as part of larger teams (conterization, APIs)• We have considerable commercial R&D experience and are
well positioned to transition any technology that is developed
PI: Vincent Emanuele, PhD
Vincent Emanuele – [email protected]
Kimia Ghobadi, Sauleh Siddiqui, Johns Hopkins University, Center for Systems Science and Engineering, Department of Civil Engineering
Teaming Overview and Objectives
Project Overview
• Existing Team: Kimia Ghobadi (inverse optimization for data analysis, real-time approximation algorithms), Sauleh Siddiqui (Bilevel optimization, machine learning in healthcare), Lauren Gardner (network analysis for data integration)
• Relevant experience: Over 50 publications in systems-level data analysis and optimization with applications to diverse healthcare settings including cancer treatment, congestive heart failure, and spread of infectious disease.
• Institutional assets: Computational capabilities.• We are seeking collaborators for TA1/TA2, specifically for
Pheonotypic Assays and real-time molecular target monitoring.
• Our team does not have unique sample/cohort access.
• We propose a systems-based computational framework using bottom-up modeling for integrating multilayered expression circuit analysis. The uniqueness of this project rests on integrating appropriate methodologies suited to expression circuits underlying each measured phenotype.
• We can contribute to expression circuit analysis and development of expression circuit analysis tools as part of TA1. We can also contribute to developing tools that will help validate the expression circuits in TA2.
Impact• Our framework bridges the gap between data-driven, ad
hoc approaches that tend to ignore the theoretical foundations of systems science and fundamental, science-based approaches that have limited empirical evidence to be broadly applicable.
• Applications to integrating multiscale, multilayered, biological datasets, which can be as broad as developing interventions to stop the spread of infectious disease.
• Metrics and milestones: data fitting using inverse optimization, high prediction accuracy using bilevel optimization, and significant measures of correlation between phenotypic observations and specific biomarkers.
• This framework will be transitioned for use by developing an open-source tool for integrating multilayered datasets.
Inve
rse
Opt
imiz
atio
n co
uple
d w
ith
Fund
amen
tal S
yste
ms
Mod
els
n
;
Sauleh Siddiqui – [email protected] – 4105166411
Machine Learning in Complex Biological Systems
Two Six Labs spans the spectrum of:
• Machine learning for biological systemsAnomaly detection from RNA-seq.Inference from high-throughput assays.de Novo protein design using neural networks.
• Machine learning on mobile platformsNeural networks on Android for real-time detection of laboratory equipment.
• Securing embedded and mobile systemsCustom operating systems to ensure data privacy and platform security.
Prime for DARPA on related programs (SHARE, Brandeis, SD2) and multiple projects for IARPA, DHS, and others.
We offer expertise applying deep learning and transfer learning across a range of bioinformatics datasets to overcome sparsity and noise.
Neural Machine Translation
Refinement with CNN
DesignConstraint
An example applying deep learning to synthetic protein design on DARPA SD2.
Two Six Labs will bring expertise in:
• Machine learning for biological systemsOvercoming sparse gene expression data.Deep learning over heterogenous data types.
• Machine learning on mobile platformsDeploying memory and power constrained neural networks on mobile devices.
• Securing embedded and mobile systemsDeploying privacy preserving embedded systems to ensure platform security.
Researched, developed and transitioned machine learning algorithms to offline mobile devices for DTRA and deployed secure mobile devices for warfighters.
Erich Izdepski, BTS Software Solutions
Teaming Overview and Objectives
Project Overview
• Team: BTS, Hippocampus Analytics• Currently providing cloud-based data search, sharing and
analysis platform to DARPA• ML and Data Science/Curation expertise; data integration• Working with various sensors: fitness type devices, RF
sensors, and IoT sensors• Our team offers our BIFROST cloud platform as an asset
for data sharing and cloud analytics. No cost for the platform, just for web hosting and software engineering operations and maintenance
• Data Compression experts with patented real-time lossless data compression algorithms (another asset)
• Our platform and capabilities can help on any TA
• Our system is aimed at enhancing collaboration to improve scientific research outcomes
• Challenges on this project:• Running ML model on a smart device and give near real-time
feedback to the user• Collecting baseline/control data and using it to identify real
signals and reduce errors• Identifying and using changes in gene expression caused by
feedback from taking device recommendations• Updating and re-deploying performance enhancing ML models in
military tactical environment
ImpactCreate a model to identify what you can be really good at based on geneticsUse the feedback device to make yourself perform betterBe happier!• We know companies already applying AI to improving
human performance• Examples:
• Mental focus, sports, musician, surgeon, pilot• Unique metrics/milestones:
• Coaching via a smartphone application. Causing positive phenotype change through change in lifestyle. Positive feedback to gene expression
Erich Izdepski, CTO of BTS – [email protected] – 703.906.5752
Human Performance Analytic Dashboard
Teaming Overview and Objectives
Project Overview
• Extensive experience turning data into information bringing together multiple techniques that outline individual adaptations and response to stressors.
• Looking for collaborators with data to validate models and methods, across new and current populations
• Collaboration with genetic and biological systems in reference to specific human muscle, tendon, ligament and bone loading
• Large Elite Athlete Performance, Physiological, Psychological, Nutritional, Medical Data Sets
• Summarized and visualized in an easy to understand format
• Complete Analytical tool that allows for Statistical, Tactical Analytics, and individualizing the interaction of all Metrics and Subjects
• Cluster, Multivariate, Archetype and Network analyses to identify variable relationships
• Identify the individual metrics that influence injury risk and optimal performance by individual and population
Impact• COMPREHENSIVE INJURY PREDICTION MODEL!• Individualized analyses in relation to performance that
identifies a performer’s strengths, weaknesses, and response metrics
• Unit analyses allows for understanding the effectiveness of each part of the unit as well as the unit as a whole.
• Analysis tool brings the communication/intervention across multidisciplinary groups
Dean Golich– [email protected]– 720-938-6066
Clinical Studies at West Point
Athletics• NCAA, Club, and Intramural athletics• All cadets participate in combatives sports and training• All cadets are required to complete routine physical fitness
assessmentsFaculty and Staff• A blend of military and government civilian• 75% Military (15% permanent, 85% rotating)• 25% Civilian (since 1994)Assessments and Predictors of Performance• Demographics• Academic Record• Physical Record• Leadership Development Record• Medical Records
Founded in 1802Undergraduate Degree • Bachelor of Science• Middle States Commission
on Higher Education • 13 Academic DepartmentsStudents• ~4,440 Aged 17-24• 75% men, 25% women• All graduates
commissioned into active duty military service
Natural ExperimentsACL Injury:
Concussion:
West Point Infrastructure Assets• Core molecular and cellular
instrument facility• Keller Army Community Hospital• John A. Feagin, Jr. Sports Medicine
Fellowship• Army Physical Therapy PhD Program
(Baylor University)• Athletic Team Trainers• Active Duty and Civilian Nurses and
Physicians
Lyme Disease:
J. Kenneth Wickiser, PhDLife Science Program DirectorAssociate ProfessorDepartment of Chemistry and Life ScienceWest Point
Contact Yuri Shoshan CEO – [email protected] - +1-312-940-5459
Augmanity Nano, Rehovot, Israel - PI: Dr. Ido Bachelet
Teaming Overview and Objectives
Measuring Biological Aptitude – TA2
Team members: • Ido Bachelet, PhD: Expert in medical DNA nanotechnology/computation, postdoc in
engineering (MIT) and DNA nanotechnology (Harvard), 15 patents, 7 on medical DNA nanotech
• Anastasia Shapiro, MSc: Expert in design and imaging of DNA devices, completing PhD on DNA-polymer computers (Technion), 6 patents on DNA nanotech
Relevant experience: • Publications in Nature Nanotechnology, Artificial Life, Nucleic Acids Res, and more
recently submitted• Patent applications on aptamer discovery platforms, DNA sensor and actuator
designs, ribosomal RNA nanodevices• Unique expertise in imaging and analyzing DNA devices in AFM Institutional assets:• Bio-AFM with live imaging and fast scan modules, next-generation sequencing core,
multi-GPU-based bioinformatics core• Affiliate at Tel Aviv Sourasky Medical Center with GMP DNA synthesis capabilities• Partnership with 2 commercial GLP animal facilities with imaging expertiseUnique access to Israel Defense Forces for samples / cohorts
• A set of novel molecular tools for real-time, in-vivo monitoring of gene and biomarker expression
• We will apply our expertise in the building of DNA-based nanoscale devices to:
• Design and manufacture in-vivo sensors triggered by expression circuit on/off state
• Develop and apply signal transduction methodologies including magnetic resonance and other technologies for external monitoring
• Technical challenges include applying contrast agents at nanoscale and design / portability of external device
Impact
Seeking partners for TA1 and device design and manufacture
Potential applications:• Biomedical imaging• Personalized medicine• Real-time treatment monitoring• Sensitive diagnosis• Gene therapy Metrics & milestones:• Design a modular DNA-based device for sensing a range
of possible biomarkers• Verify the sensor functionality in-vitro• Patent• Signal transduction experiment in-vitro• In-vivo experiments, including:
• sensor functionality• signal transduction• Stability• PK and PD• toxicity• efficiency
100 nm
Javier L. Prieto, PhD. R&D Director
Team Overview and Objectives
Project Overview
• 12 years experience deploying 85M wearable devices with physiological sensors
• Research experts on data science, artificial intelligence and algorithm development.
• Hardware experts on large scale physiological sensor design.• Access to biometrics of 25M+ active users• Biometric data: activity, HR, RHR, HRV, caloric intake, sleep• 7.5+ Billion nights of sleep logged including sleep stages• Gaming platform able to assess cognitive performance• Implemented human and AI driven coaching programs at scale• Dedicated testing facilities and experience with large scale
validation studiesSeeking collaborators to:• Acquire/Analyze genotype data relevant to the program• Acquire/Analyze blood biomarkers and proteomics data• Acquire/Analyze microbiome data• Implement sweat sensors for wearable patches for collection
of biomarker information
• Analyze continuous biometric and physiological data on a large cohort of participants among the 25M+ Fitbit users to determine correlations between these data and their mental and physical health and performance.
• Design artificial intelligence based algorithms to assess the risk and screen for stress and depression• Build coaching programs and gaming platforms to be used to guide individuals towards their peak physical, metabolic
and mental performance.• Combine biometric data and genotype data to identify and validate markers that indicate injury risk and faster recovery.• With the help of collaborators develop and deploy new connected wearable sensors for the detection of biomarkers in
sweat to help assess hydration and exertion levels.
Impact• Turnkey deployment of findings to large cohort of
users through Fitbit products and services• Development of new “wearable patch” form factor
devices for the evaluation of sweat biomarkers• Results would provide insights on:
- how activity, dietary and sleep habits affects physical and mental performance
- which genetic markers are associated with physical performance, injury risk and physical recovery
- validated sweat biomarkers for assessment of physical exertion
- how to assess mental acuity of general and aging population
- how to screen for acute stress and depression- coaching programs for glycemic control for peak
metabolic performance
Javier Prieto – [email protected] – (949) 294 0101
• Translating elite microbiomes into performance probiotics
• Mining metagenomic data for roles in energy metabolism, protein metabolism, neurology, immunology
• Discovery of a performance enhancing microbe – converts lactic acid into SCFAs to promote endurance
• Scale, clinical studies (endurance, recovery, lactate metabolism), and commercialize
• Expand cohort for novel performance microbes, metabolic pathways, markers, genes….with Syn Bio applications
• Looking for collaborations in: elite cohorts, complimentary ‘omics data & analysis, biometric detection & analysis
Jonathan Scheiman, PhDChief Executive Officer
Renee Wurth, PhDData Science Nutrition
Alex Kostic, PhDChief Technology Officer