Living systems are governed by a complex network of molecular interactions involving hundreds of thousands of interacting elements: genes, proteins, enzymes, non-coding RNA, and other signaling molecules. These systems are finely tuned to produce precise biological effects, robust enough to tolerate intrinsic and extrinsic variability, and flexible enough to adapt to environmental changes. The complex interplay of these microscopic processes drives the emergence of macroscopic phenotypic properties at the cell and organism level, and aberrations in these systems can lead to disease.

Advances in high-throughput  “omic” technologies now provide an opportunity to investigate these processes in unprecedented molecular detail, but demand the development of sophisticated computational methods to address fundamental biological complexities:

Many highly variable, interacting components

Because complex phenotypes are driven by the action of molecular systems, rather than individual genes, methods reliant upon single–marker association statistics (such as enrichment analyses) are likely to miss multi–gene effects in which no individual gene achieves independent significance. Detecting these mechanisms requires methods that treat the pathway as a whole entity, rather than the “sum of its parts”.

Heterogeneous etiologies

The molecular diversity observed in phenotypically–similar samples (e.g., cancers) suggests that there are multiple paths to an outcome. Different sets of perturbations affecting a common system may ultimately yield similar phenotypes, and causative genes may be impacted in different ways (e.g. mutation, dysregulated expression, etc.) with similar effects. Without a common signal amongst case samples, gene–level analyses can miss important disease mechanisms.

Robustness and adaptation

Living systems are robust to variability, making it necessary to distinguish tolerated alterations from those that critically impact biological processes. Likewise, because living systems are adaptive, it is necessary to distinguish sets of alterations that are compensatory from those that have a cumulative deleterious effect.

Dynamical processes

While cell signaling is a dynamic process, in most cases high-throughput data comprises only a single static snapshot or sparse time series from which dynamics must be inferred.

Our research seeks to address these challenges through the development and application of novel high-throughput analysis methods designed to elucidate systems-level differences in the structure and function of biological interaction networks associated with disease.