Peer-Reviewed Publication
Nat Biomed Eng2026;10(5):1008-1025.May 1, 2026Journal Article
Combinatorial prediction of therapeutic perturbations using causally inspired neural networks.
Guadalupe Gonzalez1,2,3, Xiang Lin4, Isuru Herath5,6,7, Kirill Veselkov1,8, Michael Bronstein9,10, Marinka Zitnik11,12,13,14
1Imperial College London, London, UK.
2F. Hoffmann-La Roche Ltd, Basel, Switzerland.
3Prescient Design, Genentech, South San Francisco, CA, USA.
4Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA.
5Merck & Co., South San Francisco, CA, USA.
6Cornell University, Ithaca, NY, USA.
7University of Pittsburgh School of Medicine-Carnegie Mellon University, Pittsburgh, PA, USA.
8Yale School of Public Health, New Haven, CT, USA.
9University of Oxford, Oxford, UK.
10AITHYRA, Vienna, Austria.
11Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA. marinka@hms.harvard.edu.
12Kempner Institute for the Study of Natural and Artificial Intelligence, Harvard University, Cambridge, MA, USA. marinka@hms.harvard.edu.
13Broad Institute of MIT and Harvard, Cambridge, MA, USA. marinka@hms.harvard.edu.
14Harvard Data Science Initiative, Cambridge, MA, USA. marinka@hms.harvard.edu.
Abstract
Phenotype-driven approaches identify disease-counteracting compounds by analysing the phenotypic signatures that distinguish diseased from healthy states. Here we introduce PDGrapher, a causally inspired graph neural network model that predicts combinatorial perturbagens (sets of therapeutic targets) capable of reversing disease phenotypes. Unlike methods that learn how perturbations alter phenoty…
Create a free account to keep reading
Free members get 10 full research views every month across publications, clinical trials, FDA clearances, adverse events, and NIH grants. No credit card required.
Want unlimited research access? See Pro plans