FuNeL: a general protocol to infer functional networks from machine learning models

Functional networks play an important role in the analysis of biological processes and systems. The inference of these networks from high-throughput (-omics) data is still an area of intense research. Similarity-based methods (e.g. gene co-expression) are the most widely used paradigms to infer networks from data. However, they define interactions only among genes that have a similar expression across different samples.

An alternative paradigm infers networks from the structure of machine learning models, as these are able to capture complex relationships between variables (genes, proteins, etc.) different from and complementary to similarity-based methods. We propose a general protocol to infer functional networks from machine learning models, called FuNeL. It assumes that genes that are used together, within a machine learning model, to classify the samples, might be also functionally related in nature. The protocol is evaluated using a test suite of 8 human cancer related datasets, and the inferred networks are compared against gene co-expression networks of equal size. The comparison highlights main differences in topology, confirms the biological relevance and complementary character of the knowledge captured by the FuNeL networks, and demonstrates their higher potential to identify known disease associations as core elements of the network.