Signatures of cell death and proliferation in perturbation transcriptomics data—from confounding factor to effective prediction

Abstract

Transcriptional perturbation signatures are valuable data sources for functional genomics. Linking perturbation signatures to screenings opens the possibility to model cellular phenotypes from expression data and to identify efficacious drugs. We linked perturbation transcriptomics data from the LINCS-L1000 project with cell viability information upon genetic (Achilles project) and chemical (CTRP screen) perturbations yielding more than 90 000 signature–viability pairs. An integrated analysis showed that the cell viability signature is a major factor underlying perturbation signatures. The signature is linked to transcription factors regulating cell death, proliferation and division time. We used the cell viability–signature relationship to predict viability from transcriptomics signatures, and identified and validated compounds that induce cell death in tumor cell lines. We showed that cellular toxicity can lead to unexpected similarity of signatures, confounding mechanism of action discovery. Consensus compound signatures predicted cell-specific drug sensitivity, even if the signature is not measured in the same cell line, and outperformed conventional drug-specific features. Our results can help in understanding mechanisms behind cell death and removing confounding factors of transcriptomic perturbation screens. To interactively browse our results and predict cell viability in new gene expression samples, we developed CEVIChE (CEll VIability Calculator from gene Expression; https://saezlab.shinyapps.io/ceviche/).

Post-perturbational transcriptional signatures of cancer cell line vulnerabilities

“…While preparing this manuscript, a similar analysis was reported investigating the patterns of association between expression and viability using several of the same datasets [24]. The authors' results are largely concordant with our findings.…”

“…However, it is not guaranteed to capture cell-context-specific effects, and only a portion of these effects are likely to be associated with the resulting cell death or cell cycle arrest processes. By integrating measured viability data, our modeling approach explicitly decomposes the transcriptional response to a perturbation into a viability-related component (which could suggest a drug's killing mechanism), and a viability-independent component (which could better resolve the drug's MOA, as was recently shown [24].…”

“…A number of studies have presented broad characterizations of the expression-viability relationship by examining responses to drugs with various mechanisms, revealing patterns in gene expression that are associated with fitness loss [18,24]. Another line of work has focused on building predictive models to reliably estimate a cell's future viability response given the transcriptional profile [20].…”

Modeling drug mechanism of action with large scale gene-expression profiles using GPAR, an artificial intelligence platform

“…As a matter of fact, a wide variety of machine learning methods have been developed to help to understand the mechanisms underlying gene expression [ 1 , 16 ], and thus far some of previous published work has demonstrated that deep learning is an effective approach to connect signatures to prior knowledge such as side effects, indication, targets or drug sensitivity [ 1 , 16 – 19 ], and has Noted that the high hidden layer feature of deep learning could effectively reduce the batch effect [ 15 ]. Considering the usage of the existing tools limited by the high cost in computational resource and the difficulty in model training and accuracy evaluation, here we report a new efficient method for querying gene expression data by taking advantage of deep neural networks to learn an embedding and do more precision classification.…”

Drug-induced cell viability prediction from LINCS-L1000 through WRFEN-XGBoost algorithm

“…Similarly, Szalai Bence et al conducted a model prediction analysis based on the correlation between the differentially expressed genes measured in the cell lines and the drug sensitivity under the action of the the drug at a specific concentration, and found that the cell line response was correlated with the drug concentration and time. However, the model achieved low accuracy and poor fitting in the prediction process because it ignored the non-linear characteristics between differentially expressed genes and the drug sensitivity [ 11 ].…”