Advancing Epigenetic Predictors with Scalable Machine Learning
summary
Company converts Illumina methylation arrays into validated health predictors using R (Sesame, MinFi), Python/PyTorch and Posit Workbench on AWS (EFS, S3, EC2, SageMaker) with parallelization and sharding to cut training from weeks to hours. A model sweep starts at 5,000 models reduced to 1,700 after validation, saving almost a year of product development, producing 30+ publications in four years, and shipping consumer reports plus an internal Shiny visualization dashboard.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:04Epigenetic testing advantages
- 3:34DNA methylation and environment
- 6:01Scalable ML in epigenetics
- 8:06HPC to Posit Workbench
- 10:18Posit Workbench benefits
- 13:31Posit Workbench use cases
- 16:42Empowering non-technical teams
- 17:24Future plans and AI
- 18:26Q&A
related talks
26mShiva Amiri, CEO, BioSymetrics at MLconf Seattle 2017
42mHow Novel Compute Technology Transforms Life Science Research
22mAccelerating ML impact
29mDeveloping and Delivering Personalized Polygenic Scores at Scale
45mDigital Disruption in health & medical research
35mFrom cells to drug responses - machine learning in cancer research
26mAccelerating innovation in the healthcare and life sciences industry
31mWhole Genome Methylation Patterns as a Biomarkers for EHR Imputation