Predicting cellular responses to complex perturbations in high‐throughput screens

Abstract Recent advances in multiplexed single‐cell transcriptomics experiments facilitate the high‐throughput study of drug and genetic perturbations. However, an exhaustive exploration of the combinatorial perturbation space is experimentally unfeasible. Therefore, computational methods are needed...

Full description

Saved in:
Bibliographic Details
Main Authors: Mohammad Lotfollahi, Anna Klimovskaia Susmelj, Carlo De Donno, Leon Hetzel, Yuge Ji, Ignacio L Ibarra, Sanjay R Srivatsan, Mohsen Naghipourfar, Riza M Daza, Beth Martin, Jay Shendure, Jose L McFaline‐Figueroa, Pierre Boyeau, F Alexander Wolf, Nafissa Yakubova, Stephan Günnemann, Cole Trapnell, David Lopez‐Paz, Fabian J Theis
Format: Article
Language:English
Published: Springer Nature 2023-05-01
Series:Molecular Systems Biology
Subjects:
Online Access:https://doi.org/10.15252/msb.202211517
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849235771251228672
author Mohammad Lotfollahi
Anna Klimovskaia Susmelj
Carlo De Donno
Leon Hetzel
Yuge Ji
Ignacio L Ibarra
Sanjay R Srivatsan
Mohsen Naghipourfar
Riza M Daza
Beth Martin
Jay Shendure
Jose L McFaline‐Figueroa
Pierre Boyeau
F Alexander Wolf
Nafissa Yakubova
Stephan Günnemann
Cole Trapnell
David Lopez‐Paz
Fabian J Theis
author_facet Mohammad Lotfollahi
Anna Klimovskaia Susmelj
Carlo De Donno
Leon Hetzel
Yuge Ji
Ignacio L Ibarra
Sanjay R Srivatsan
Mohsen Naghipourfar
Riza M Daza
Beth Martin
Jay Shendure
Jose L McFaline‐Figueroa
Pierre Boyeau
F Alexander Wolf
Nafissa Yakubova
Stephan Günnemann
Cole Trapnell
David Lopez‐Paz
Fabian J Theis
author_sort Mohammad Lotfollahi
collection DOAJ
description Abstract Recent advances in multiplexed single‐cell transcriptomics experiments facilitate the high‐throughput study of drug and genetic perturbations. However, an exhaustive exploration of the combinatorial perturbation space is experimentally unfeasible. Therefore, computational methods are needed to predict, interpret, and prioritize perturbations. Here, we present the compositional perturbation autoencoder (CPA), which combines the interpretability of linear models with the flexibility of deep‐learning approaches for single‐cell response modeling. CPA learns to in silico predict transcriptional perturbation response at the single‐cell level for unseen dosages, cell types, time points, and species. Using newly generated single‐cell drug combination data, we validate that CPA can predict unseen drug combinations while outperforming baseline models. Additionally, the architecture's modularity enables incorporating the chemical representation of the drugs, allowing the prediction of cellular response to completely unseen drugs. Furthermore, CPA is also applicable to genetic combinatorial screens. We demonstrate this by imputing in silico 5,329 missing combinations (97.6% of all possibilities) in a single‐cell Perturb‐seq experiment with diverse genetic interactions. We envision CPA will facilitate efficient experimental design and hypothesis generation by enabling in silico response prediction at the single‐cell level and thus accelerate therapeutic applications using single‐cell technologies.
format Article
id doaj-art-418ea36376f0413cb228551ff196f2db
institution Kabale University
issn 1744-4292
language English
publishDate 2023-05-01
publisher Springer Nature
record_format Article
series Molecular Systems Biology
spelling doaj-art-418ea36376f0413cb228551ff196f2db2025-08-20T04:02:41ZengSpringer NatureMolecular Systems Biology1744-42922023-05-0119611910.15252/msb.202211517Predicting cellular responses to complex perturbations in high‐throughput screensMohammad Lotfollahi0Anna Klimovskaia Susmelj1Carlo De Donno2Leon Hetzel3Yuge Ji4Ignacio L Ibarra5Sanjay R Srivatsan6Mohsen Naghipourfar7Riza M Daza8Beth Martin9Jay Shendure10Jose L McFaline‐Figueroa11Pierre Boyeau12F Alexander Wolf13Nafissa Yakubova14Stephan Günnemann15Cole Trapnell16David Lopez‐Paz17Fabian J Theis18Helmholtz Center Munich – German Research Center for Environmental Health, Institute of Computational BiologyMeta AIHelmholtz Center Munich – German Research Center for Environmental Health, Institute of Computational BiologyHelmholtz Center Munich – German Research Center for Environmental Health, Institute of Computational BiologyHelmholtz Center Munich – German Research Center for Environmental Health, Institute of Computational BiologyHelmholtz Center Munich – German Research Center for Environmental Health, Institute of Computational BiologyDepartment of Genome Sciences, University of WashingtonDepartment of Bioengineering, University of CaliforniaDepartment of Genome Sciences, University of WashingtonDepartment of Genome Sciences, University of WashingtonDepartment of Genome Sciences, University of WashingtonDepartment of Biomedical Engineering, Columbia UniversityDepartment of Electrical Engineering and Computer Sciences, University of CaliforniaHelmholtz Center Munich – German Research Center for Environmental Health, Institute of Computational BiologyMeta AIDepartment of Computer Science, Technical University of MunichDepartment of Genome Sciences, University of WashingtonWellcome Trust Sanger Institute, Wellcome Genome Campus, HinxtonHelmholtz Center Munich – German Research Center for Environmental Health, Institute of Computational BiologyAbstract Recent advances in multiplexed single‐cell transcriptomics experiments facilitate the high‐throughput study of drug and genetic perturbations. However, an exhaustive exploration of the combinatorial perturbation space is experimentally unfeasible. Therefore, computational methods are needed to predict, interpret, and prioritize perturbations. Here, we present the compositional perturbation autoencoder (CPA), which combines the interpretability of linear models with the flexibility of deep‐learning approaches for single‐cell response modeling. CPA learns to in silico predict transcriptional perturbation response at the single‐cell level for unseen dosages, cell types, time points, and species. Using newly generated single‐cell drug combination data, we validate that CPA can predict unseen drug combinations while outperforming baseline models. Additionally, the architecture's modularity enables incorporating the chemical representation of the drugs, allowing the prediction of cellular response to completely unseen drugs. Furthermore, CPA is also applicable to genetic combinatorial screens. We demonstrate this by imputing in silico 5,329 missing combinations (97.6% of all possibilities) in a single‐cell Perturb‐seq experiment with diverse genetic interactions. We envision CPA will facilitate efficient experimental design and hypothesis generation by enabling in silico response prediction at the single‐cell level and thus accelerate therapeutic applications using single‐cell technologies.https://doi.org/10.15252/msb.202211517generative modelinghigh‐throughput screeningmachine learningperturbation predictionsingle‐cell transcriptomics
spellingShingle Mohammad Lotfollahi
Anna Klimovskaia Susmelj
Carlo De Donno
Leon Hetzel
Yuge Ji
Ignacio L Ibarra
Sanjay R Srivatsan
Mohsen Naghipourfar
Riza M Daza
Beth Martin
Jay Shendure
Jose L McFaline‐Figueroa
Pierre Boyeau
F Alexander Wolf
Nafissa Yakubova
Stephan Günnemann
Cole Trapnell
David Lopez‐Paz
Fabian J Theis
Predicting cellular responses to complex perturbations in high‐throughput screens
Molecular Systems Biology
generative modeling
high‐throughput screening
machine learning
perturbation prediction
single‐cell transcriptomics
title Predicting cellular responses to complex perturbations in high‐throughput screens
title_full Predicting cellular responses to complex perturbations in high‐throughput screens
title_fullStr Predicting cellular responses to complex perturbations in high‐throughput screens
title_full_unstemmed Predicting cellular responses to complex perturbations in high‐throughput screens
title_short Predicting cellular responses to complex perturbations in high‐throughput screens
title_sort predicting cellular responses to complex perturbations in high throughput screens
topic generative modeling
high‐throughput screening
machine learning
perturbation prediction
single‐cell transcriptomics
url https://doi.org/10.15252/msb.202211517
work_keys_str_mv AT mohammadlotfollahi predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT annaklimovskaiasusmelj predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT carlodedonno predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT leonhetzel predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT yugeji predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT ignaciolibarra predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT sanjayrsrivatsan predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT mohsennaghipourfar predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT rizamdaza predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT bethmartin predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT jayshendure predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT joselmcfalinefigueroa predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT pierreboyeau predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT falexanderwolf predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT nafissayakubova predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT stephangunnemann predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT coletrapnell predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT davidlopezpaz predictingcellularresponsestocomplexperturbationsinhighthroughputscreens
AT fabianjtheis predictingcellularresponsestocomplexperturbationsinhighthroughputscreens