The Compositional Perturbation Autoencoder (CPA) is a deep generative framework to learn effects of perturbations at the single-cell level. CPA performs OOD predictions of unseen combinations of drugs, learns interpretable embeddings, estimates dose-response curves, and provides uncertainty estimates.
License: MIT License
Shell 0.02%Python 0.43%Jupyter Notebook 99.56%
compositperterbanalysis's Introduction
About Me
๐ Hi, Iโm @BrianEads. I'm currently doing cloud engineering for small molecules R&D for Bayer Crop Science. My background is in data science, genomics and analytics engineering.
Professional Background
๐ Iโm interested in data, machine learning and automation. My team deploys a wide array of tools in this space designed to accelerate discovery of novel small molecules, and to augment human decisions with new analytics pipelines to speed time to market.
I have a background in molecular biology and biochemistry, extensive experience in genomics and bioinformatics, and an abiding interest in using machine learning for computational biology domains. Much of our data engineering work straddles on-prem and cloud assets as we move further into new patterns of GCP-stored data (accessed via BigQuery or API). Our cloud expertise is heavily AWS and we use GitHub (natch) and other enterprise-supported platforms across the DevOps lifecycle of our products.
Problem Domains
๐ฑ My team is currently leveraging serverless patterns for our workloads, and exploring emerging patterns like deep learning and transfer learning. In machine learning spaces, we help our modeler colleagues to automate model deployment and monitoring.
Side Projects
๐๏ธ I enjoy several tech-related hobbies, including archaeogenomics and Rasp-pi hacking.