Biological Cartography: Building and Benchmarking Representations of Life
Purpose: By the end of the workshop, participants should have a good understanding of what maps of biology are, and how to build and benchmark them using image-based profiles.
Presentation: Participants will be introduced to Recursion, its mission, and its approach to industrializing drug discovery. They will learn how Recursion builds and benchmarks maps of biology using image-based profiles (as outlined in this paper co-authored by Recursion and Genentech), and uses these maps for drug discovery. Along the way, participants will gain an understanding of what an EFAAR pipeline is, as well as discuss the issues one should consider when configuring these pipelines for different applications. Existing implementations of EFAAR pipelines will be used as motivating examples. Participants will also be introduced to a suite of metrics that can be used to benchmark and compare different maps of biology.
Demo: Using a publicly-available notebook, participants will follow along as they learn to build and benchmark maps of biology on their own, using embeddings from the public image sets RxRx3 and JUMP-CP. The notebook will use code from the public EFAAR_benchmarking repo.
Unbiased morphological profiling: state-of-the-art applications and emerging frontiers
Format
Intro
Panel discussion outline
Ghost Cytometry: An AI-ML technology for High Resolution, Label-free Cellular Fingerprinting
Ghost Cytometry technology is an AI-based hi-res label-free cell analysis and sorting technology for cell therapeutic development and drug discovery providing an entirely new window into Cellular Fingerprinting. In this workshop, we will cover how the technology enables detection of intra-cellular morphological structures and restructuring. The VisionSort System is the first instrument that allows the physical analysis and sorting of cells based on their morphology, cytoskeletal structure using AI-based technology. With > 9 million features of high dimensional information per second, VisionSort can profile intracellular structures and spatial information across a wide range of applications.