Sparse Regression Based Structure Learning of Stochastic Reaction Networks from Single Cell Snapshot Time Series.

摘要: Stochastic chemical reaction networks constitute a model class to quantitatively describe dynamics and cell-to-cell variability in biological systems. The topology of these typically is only partially characterized due experimental limitations. Current approaches for refining network are based on the explicit enumeration alternative topologies therefore restricted small problem instances with almost complete knowledge. We propose reactionet lasso, computational procedure that derives stepwise sparse regression approach basis Chemical Master Equation, enabling large-scale structure learning by implicitly accounting billions variants. have assessed capabilities lasso synthetic data TRAIL induced apoptosis signaling cascade comprising 70 reactions. find able efficiently recover systems, ab initio, high sensitivity specificity. With 6000 possible reactions over 102000 topologies. In conjunction information rich single cell technologies such as RNA sequencing or mass cytometry, will enable learning, particularly areas partial knowledge, cancer biology, thereby detection pathological alterations networks. provide software allow wide applicability lasso.