Computational translation of genomic responses from experimental model systems to humans

Abstract

The high failure rate of therapeutics showing promise in mouse models to translate to patients is a pressing challenge in biomedical science. Though retrospective studies have examined the fidelity of mouse models to their respective human conditions, approaches for prospective translation of insights from mouse models to patients remain relatively unexplored. Here, we develop a semi-supervised learning approach for inference of disease-associated human differentially expressed genes and pathways from mouse model experiments. We examined 36 transcriptomic case studies where comparable phenotypes were available for mouse and human inflammatory diseases and assessed multiple computational approaches for inferring human biology from mouse datasets. We found that semi-supervised training of a neural network identified significantly more true human biological associations than interpreting mouse experiments directly. Evaluating the experimental design of mouse experiments where our model was most successful revealed principles of experimental design that may improve translational performance. Our study shows that when prospectively evaluating biological associations in mouse studies, semi-supervised learning approaches, combining mouse and human data for biological inference, provide the most accurate assessment of human in vivo disease processes. Finally, we proffer a delineation of four categories of model system-to-human “Translation Problems” defined by the resolution and coverage of the datasets available for molecular insight translation and suggest that the task of translating insights from model systems to human disease contexts may be better accomplished by a combination of translation-minded experimental design and computational approaches.

Transfer of regulatory knowledge from human to mouse for functional genomics analysis

“…1A). Recent studies indicate that footprints outperform mapping gene sets (Schubert et al , 2018;Cantini et al , 2018) . Since most of these footprints are generated for the application in humans, their usability in model organisms is uncertain.…”

“…This question is of importance since the study of human diseases is limited by the availability of patient data and ethical concerns, and are often complemented with experimental work in model organisms, in particular mice ( Mus Musculus ;The Mouse in Biomedical Research, 2007) . Disease alterations of gene expression in human can be estimated from mouse transcriptomic data (Normand et al , 2018;Brubaker et al , 2019) . Furthermore, previous studies suggest that pathway and TF footprints are evolutionarily conserved between mice and humans: pathway footprints derived from mouse B cells can provide valuable insights into human cancer (Tenenbaum et al , 2008) , and inferred prostate-specific gene regulatory networks of mice and humans overlap in over 70 % (Aytes et al , 2014) .…”

Predicting transcriptional responses to cold stress across plant species

“…Here, we demonstrate that, even though orthology is not an effective predictor of transcriptional responses to cold stress across even closely related species, it is possible to train supervised classification models using data from one species to predict which genes will respond to cold stress in another species. The usefulness of supervised classification algorithms has been demonstrated for a range of biological applications, such as distinguishing gene models with the potential for expression ( 14 ), inferring human gene expression based on a mouse model ( 15 ), predicting functional annotations of individual gene models from functional genomic data ( 16 ), distinguishing genes involved in specialized or primary metabolism ( 17 ), and predicting posttranslational modification sites ( 18 ). In this study, we generated transcriptional data from four closely related species: foxtail millet, pearl millet, switchgrass, and proso millet ( Fig.…”

Cancer systems immunology

“…Computational biologists have developed a variety of tools to translate findings from preclinical models to humans when simple matching of orthologs is insufficient for predicting responses ( Brubaker and Lauffenburger, 2020 ). To overcome differences in gene-to-function relationships and predict human responses from rodent data, researchers used Bayesian analysis of gene expression to define ‘functional orthologs’ across species ( Chikina and Troyanskaya, 2011 ), and others have applied unsupervised and semi-supervised machine learning approaches to transcriptomic and proteomic data generated in rodent models to predict human responses ( Brubaker et al, 2019 ). On collaborative initiatiative, termed SBV-IMPROVER (Systems Biology Verification for Industrial Methodology for PROcess VErification in Research), sought to develop computational methods capable of cross-species translation using multimodal datasets including transcriptomics, phosphoproteomics, and cytokine data ( Poussin et al, 2014 ; Rhrissorrakrai et al, 2015 ).…”