Timezone: »

 
Poster
Predicting Cellular Responses to Novel Drug Perturbations at a Single-Cell Resolution
Leon Hetzel · Simon Boehm · Niki Kilbertus · Stephan Günnemann · mohammad lotfollahi · Fabian Theis

Tue Nov 29 09:00 AM -- 11:00 AM (PST) @ Hall J #524
in Poster Session 1 »

Single-cell transcriptomics enabled the study of cellular heterogeneity in response to perturbations at the resolution of individual cells. However, scaling high-throughput screens (HTSs) to measure cellular responses for many drugs remains a challenge due to technical limitations and, more importantly, the cost of such multiplexed experiments. Thus, transferring information from routinely performed bulk RNA HTS is required to enrich single-cell data meaningfully.We introduce chemCPA, a new encoder-decoder architecture to study the perturbational effects of unseen drugs. We combine the model with an architecture surgery for transfer learning and demonstrate how training on existing bulk RNA HTS datasets can improve generalisation performance. Better generalisation reduces the need for extensive and costly screens at single-cell resolution. We envision that our proposed method will facilitate more efficient experiment designs through its ability to generate in-silico hypotheses, ultimately accelerating drug discovery.

Keywords: Drug Discovery · transfer learning · Disentanglement · unsupervised · genomics · single cell · perturbation
[ Paper]  [ Poster]  [ OpenReview] 

Author Information

Leon Hetzel (TUM, Helmholtz Munich)
Simon Boehm (Swiss Federal Institute of Technology)
Simon Boehm

Earlier this year, I finished my Data Science MSc at ETH Zurich. My Master's thesis was supervised by Prof. Theis and Prof. Kilbertus at Helmholtz Munich, in the area of computation biology. Currently, I'm more interested in DL systems research and am working on projects around [reimplementing 3D-parallel distributed deep learning](https://github.com/siboehm/shallowspeed) and [LLVM compilers for tree-based ML models](https://github.com/siboehm/lleaves). Previously I was a compiler engineering intern at AMD, working on the team that builds the MLIR compiler for a dataflow-centric deep learning accelerator (Xilinx AI Engines).

Niki Kilbertus (TUM & Helmholtz AI)
Stephan Günnemann (Technical University of Munich)
mohammad lotfollahi (Helmholtz Zentrum München)
Fabian Theis (Helmholtz Munich)

More from the Same Authors