PuSH - Publikationsserver des Helmholtz Zentrums München

Szałata, A. ; Benz, A.* ; Cannoodt, R.* ; Cortes, M.* ; Fong, J.* ; Kuppasani, S.* ; Lieberman, R.* ; Liu, T.* ; Mas-Rosario, J.A.* ; Meinl, R.* ; Nourisa, J.* ; Tumiel, R.* ; Tunjic, T.M.* ; Wang, M.* ; Weber, N.* ; Zhao, H.* ; Anchang, B.* ; Theis, F.J. ; Luecken, M. ; Burkhardt, D.B.*

A benchmark for prediction of transcriptomic responses to chemical perturbations across cell types.

In: (38th Conference on Neural Information Processing Systems (NeurIPS 2024), 9-15 December 2024, Vancouver). 2024.
Postprint
ingle-cell transcriptomics has revolutionized our understanding of cellular hetero-1 geneity and drug perturbation effects. However, its high cost and the vast chemical2 space of potential drugs present barriers to experimentally characterizing the effect3 of chemical perturbations in all the myriad cell types of the human body. To4 overcome these limitations, several groups have proposed using machine learning5 methods to directly predict the effect of chemical perturbations either across cell6 contexts or chemical space. However, advances in this field have been hindered7 by a lack of well-designed evaluation datasets and benchmarks. To drive innova-8 tion in perturbation modeling, the Open Problems Perturbation Prediction (OP3)9 benchmark introduces a framework for predicting the effects of small molecule per-10 turbations on cell type-specific gene expression. OP3 leverages the Open Problems11 in Single-cell Analysis benchmarking infrastructure and is enabled by a new single-12 cell perturbation dataset, encompassing 146 compounds tested on human blood13 cells. The benchmark includes diverse data representations, evaluation metrics,14 and winning methods from our “Single-cell perturbation prediction: generaliz-15 ing experimental interventions to unseen contexts“ competition at NeurIPS 2023.16 We envision that the OP3 benchmark and competition will drive innovation in17 single-cell perturbation prediction by improving the accessibility, visibility, and18 feasibility of this challenge, thereby promoting the impact of machine learning in19 drug discovery.
Weitere Metriken?
[➜Einloggen]
Zusatzinfos bearbeiten [➜Einloggen]
Publikationstyp Artikel: Konferenzbeitrag
Konferenztitel 38th Conference on Neural Information Processing Systems (NeurIPS 2024)
Konferzenzdatum 9-15 December 2024
Konferenzort Vancouver
Institut(e) Institute of Computational Biology (ICB)