Predicting multicellular function through multi-layer tissue networks

Abstract

MotivationUnderstanding functions of proteins in specific human tissues is essential for insights into disease diagnostics and therapeutics, yet prediction of tissue-specific cellular function remains a critical challenge for biomedicine.ResultsHere, we present OhmNet, a hierarchy-aware unsupervised node feature learning approach for multi-layer networks. We build a multi-layer network, where each layer represents molecular interactions in a different human tissue. OhmNet then automatically learns a mapping of proteins, represented as nodes, to a neural embedding-based low-dimensional space of features. OhmNet encourages sharing of similar features among proteins with similar network neighborhoods and among proteins activated in similar tissues. The algorithm generalizes prior work, which generally ignores relationships between tissues, by modeling tissue organization with a rich multiscale tissue hierarchy. We use OhmNet to study multicellular function in a multi-layer protein interaction network of 107 human tissues. In 48 tissues with known tissue-specific cellular functions, OhmNet provides more accurate predictions of cellular function than alternative approaches, and also generates more accurate hypotheses about tissue-specific protein actions. We show that taking into account the tissue hierarchy leads to improved predictive power. Remarkably, we also demonstrate that it is possible to leverage the tissue hierarchy in order to effectively transfer cellular functions to a functionally uncharacterized tissue. Overall, OhmNet moves from flat networks to multiscale models able to predict a range of phenotypes spanning cellular subsystems. Availability and implementationSource code and datasets are available at http://snap.stanford.edu/ohmnet.

Supervised-learning is an accurate method for network-based gene classification

“…Akin to LP [38,75,76] , node embeddings also offer a convenient route to incorporating multiple networks into SL approaches. While methods such as SL-I and SL-A may require concatenating the original networks or integrating them into a single network before learning, recent work has shown that SL-E-based methods can embed information from multiple molecular/heterogeneous networks and learn gene classifiers in tandem [77][78][79][80][81][82][83][84][85] . However, none of these studies have compared the variety of SL-E methods to learning directly on the adjacency matrix.…”

Multilayer network modeling of integrated biological systems

“…To this aim, multiples sources of omics data encode different layers, representing a biological system as a network of networks. This integrated perspective allows for more predictive performances [17,18,19] and has been shown to better characterize the evolution of complex diseases such as cancer [20], as well as to better understand the response to genetic and metabolic perturbations in complex organisms like E. coli [21].…”

Formatting biological big data for modern machine learning in drug discovery

“…A successful and intuitive application of multilayer gene networks is PARADIGM, a system that models the central dogma of biology (DNA–mRNA–protein) with multiple patient‐specific “omics” measurements, and uses probabilistic inference to identify altered protein activities in each patient. Another application regards the rewiring of protein–protein interactions in 107 human tissues by means of a multilevel interactome that was shown to capture tissue‐specific functions of the proteins …”