Machine learning ranking of plausible (un)explored synergistic gene combinations using sensitivity indices of time series measurements of Wnt signaling pathway.

Abstract

Combinations of genes or proteins work in synergy at different times and durations in a signaling pathway. However, which combinations are prevalent at a particular time point or duration is mostly not known. Sensitivity analysis plays a major role in computing the strength of the influence of involved factors in any phenomena under investigation. When applied to expression profiles of various intra/extracellular factors that work in a signaling pathway, the variance- and density-based analysis yields a range of sensitivity indices for individual and various combinations of factors. These combinations denote the higher order interactions among the involved factors, which might be of interest. In this work, after estimating the individual effects of factors for a higher order combination, the individual indices are considered as discriminative features. Exploiting the analogy of prioritizing webpages using ranking algorithms, for a particular order, a full set of combinations of genes can be prioritized based on these features using a powerful support vector ranking algorithm. Recording the changing rankings of the combinations over time points and durations reveals which higher order combinations influence the pathway and when and where an intervention might be necessary to affect the pathway. Integration, innovation, and insight Combinations of genes or proteins work in synergy at different times and durations in a signaling pathway. However, which combinations are prevalent at a particular time point or duration is mostly not known. This work develops a search engine that reveals ground-breaking results in the form of higher order (un)explored/(un)tested combinations (as biological hypotheses), based on sensitivity indices. These indices capture the strength of influence of factors (here genes/proteins) that affect a signaling pathway. Recording the changing rankings of these combinations over time points and durations reveals how higher order combinations behave within the pathway. Significance The manuscript develops a search engine that reveals ground-breaking results in the form of higher order (un)explored/(un)tested combinations of genes/proteins (as biological hypotheses), based on sensitivity indices that capture the strength of influence of factors (here genes/proteins) that affect the Wnt signaling pathway. The pipeline uses kernel-based sensitivity indices to capture the influence of the factors in a pathway and employs powerful support vector ranking algorithm. Because of the above point, biologists/oncologists will be able to narrow down their search to particular combinations that are ranked and, if a synergistic functioning is confirmed, will be able to study the mechanism between the components of a combination, in the Wnt pathway. The search engine design is not only limited to one dataset and a range of combinations of genes/proteins. The framework can be applied/modified to all problems where one is interested in searching for particular combinations of factors involved in a particular phenomena. Recording the changing rankings of the combinations over time points and durations reveals how higher order interactions behave within the pathway and when and where an intervention might be necessary to influence the pathway, for therapeutic purpose. It reveals the various unexplored FZD-WNT combinations that have been untested till now in the Wnt pathway.