During the last two decades, molecular biology has undergone a revolution in high-dimensional measurement techniques.  The measured data reflects multi-scale relationships in biology - correlations arising from cell admixture, across and within pathways, regulator to regulon, and so on.  These broad, weak correlations obscure signal, while biological and platform noise confound it.  Telling correlation from causality is thus harder than ever, while high-throughput intervention-based experimental designs are frequently impractical or unethical.  Computational methods must leverage cross-sectional data to infer causes of phenotypes and to predict actionable targets of interventions.  Suitable approaches have existed for decades, but computational complexity has limited their applicability at the scale of current and especially nascent technologies.  GNS Healthcare has implemented, and productized a causal inference platform, REFS(tm), equal to the scale of the inference problem.  Drawing inspiration from statistical mechanics, GNS took the stochastic ensemble approach to inference, alleviating the need to find a precise solution in a combinatorial search space.  REFS can tackle the problem of inferring the seating in the molecular game of telephone and find cause-effect relationships while controlling for confounder variables.  We illustrate the capabilities of our platform using a recent study of causes of coronary artery disease.

Molecular biology meets big data: Winning the game of telephone

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.