Percolate: an exponential family JIVE model to design DNA-based predictors of drug response

Abstract

Motivation: Anti-cancer drugs may elicit resistance or sensitivity through mechanisms which involve several genomic layers. Nevertheless, we have demonstrated that gene expression contains most of the predictive capacity compared to the remaining omic data types. Unfortunately, this comes at a price: gene expression biomarkers are often hard to interpret and show poor robustness. Results: To capture the best of both worlds, i.e. the accuracy of gene expression and the robustness of other genomic levels, such as mutations, copy-number or methylation, we developed Percolate, a computational approach which extracts the joint signal between gene expression and the other omic data types. We developed an out-of-sample extension of Percolate which allows predictions on unseen samples without the necessity to recompute the joint signal on all data. We employed Percolate to extract the joint signal between gene expression and either mutations, copy-number or methylation, and used the out-of sample extension to perform response prediction on unseen samples. We showed that the joint signal recapitulates, and sometimes exceeds, the predictive performance achieved with each data type individually. Importantly, molecular signatures created by Percolate do not require gene expression to be evaluated, rendering them suitable to clinical applications where only one data type is available.

An in-depth comparison of linear and non-linear joint embedding methods for bulk and single-cell multi-omics

“…On the other hand, these methods underperformed with respect to baseline non-linear methods in terms of generation coherence. Regarding downstream supervised tasks, we observed that joint embedding can lead to improved performance when only a single modality is available in the test data, which verifies previous results for linear methods [27]. If data from both modalities are available at test time, joint embedding did not provide a significant advantage in the tasks that we tested here.…”

An in-depth comparison of linear and non-linear joint embedding methods for bulk and single-cell multi-omics

“…On the other hand, these methods underperformed with respect to baseline non-linear methods in terms of generation coherence. Regarding downstream supervized tasks, we observed that joint embedding can lead to improved performance when only a single modality is available in the test data, which verifies previous results for linear methods [ 29 ]. If data from both modalities are available at test time, joint embedding did not provide a significant advantage in the tasks that we tested here.…”

User-friendly exploration of epigenomic data in single cells using sincei

“…sincei accommodates these popular choices through the log1p-PCA, LSA, and LDA methods. Furthermore, we also offer a scalable implementation of GLM-PCA 12,13 , a flexible dimensionality reduction method that can accommodate a broad range of signal types, such as counts (Poisson), binary (Bernoulli), ratio (Beta). Developers can easily extend this list, providing users with the flexibility needed to explore diverse statistical modeling options, particularly instrumental in developing novel assays.…”