Using ILP to Identify Pathway Activation Patterns in Systems Biology

Abstract

We show a logical aggregation method that, combined with propositionalization methods, can construct novel structured biological features from gene expression data. We do this to gain understanding of pathway mechanisms, for instance, those associated with a particular disease. We illustrate this method on the task of distinguishing between two types of lung cancer; Squamous Cell Carcinoma (SCC) and Adenocarcinoma (AC). We identify pathway activation patterns in pathways previously implicated in the development of cancers. Our method identified a model with comparable predictive performance to the winning algorithm of a recent challenge, while providing biologically relevant explanations that may be useful to a biologist.

Keywords: Barcode; Biological pathways; Logical aggregation; Reactome; TreeLiker; Warmr.

Fig. 1

Method overview

Fig. 2

Reaction Graph illustrations. There are three types of relationships between reactions: follows (black solid lines), activation (green dashed), and inhibition (red dash-dotted).

Fig. 3

Illustration of logical aggregation. Known biological mechanisms can be represented as OR or AND gates. The triangular nodes are binary probe values, created using barcode.

Fig. 4

Results (a) Example of the hexose uptake pathway for a particular individual. Green squares: on reactions, red octagons: off reactions. Identified feature of three on reactions shown in pink, dashed box. (b) ROC curves comparing performance. SBV Improver: blue, dashed; TreeLiker & J48: green, solid; Warmr & OneR: red, dash-dotted.

Fig. 5

The pattern found by Warmr instantiated for individual GSM1065725. There is a self-activating loop, highlighted by the grey box.

Fig. 6

The three features in the subgroup description. Solid lines represent the feature found by TreeLiker, dotted lines show the Warmr extensions. on reactions: green, rounded squares; off: red octagons; on or off: blue squares.