A Multi-Species Functional Embedding Integrating Sequence and Network Structure

Abstract

Abstract.A key challenge to transferring knowledge between species is that different species have fundamentally different genetic architectures. Initial computational approaches to transfer knowledge across species have relied on measures of heredity such as genetic homology, but these approaches suffer from limitations. First, only a small subset of genes have homologs, limiting the amount of knowledge that can be transferred, and second, genes change or repurpose functions, complicating the transfer of knowledge. Many approaches address this problem by expanding the notion of homology by leveraging high-throughput genomic and proteomic measurements, such as through network alignment.In this work, we take a new approach to transferring knowledge across species by expanding the notion of homology through explicit measures of functional similarity between proteins in different species. Specifically, our kernel-based method, Handl (Homology Assessment across Networks using Diffusion and Landmarks), integrates sequence and network structure to create a functional embedding in which proteins from different species are embedded in the same vector space. We show that inner products in this space and the vectors themselves capture functional similarity across species, and are useful for a variety of functional tasks. We perform the first whole-genome method for predicting phenologs, generating many that were previously identified, but also predicting new phenologs supported from the biological literature. We also demonstrate the Handl embedding captures pairwise gene function, in that gene pairs with synthetic lethal interactions are significantly separated in Handl space, and the direction of separation is conserved across species. Software for the Handl algorithm is available at http://bit.ly/lrgr-handl.

To Embed or Not: Network Embedding as a Paradigm in Computational Biology

“…A recent work by Fan et al (2017) uses an embedding-based approach, MuNK, to compare networks across species by assessing similarity via embedded network topologies. The idea is to project the nodes of the two networks into the same Euclidean space in a way that preserves their intra-species network similarity and inter-species sequence similarity.…”

“…For each species separately, a kernel similarity function is defined, and the corresponding embedding is computed by matrix decomposition. To tie the projections together, Fan et al (2017) assume a given set of known matches, regarded as landmarks, between the two networks. A similar embedding approach that does not require a known subset of correspondences was suggested in (Heimann et al, 2018).…”

“…The interpretability of the embedding space is one of the primary benefits of embedding techniques over standard approaches in the case of network alignment. For example, the embedding space learned by MuNK captures biological information beyond pairwise node alignment, specifically, cross-species synthetic lethal interactions (Fan et al, 2017).…”

“…The integration challenge is also tackled by (Gligorijevic et al, 2018) who learn a compact node representation using deep autoencoders. In Fan et al (2017), the cross-species embedding is utilized to infer protein function. Zitnik and Leskovec (2017) suggest a network embedding approach for predicting tissue-specific protein function, which encourages proteins to share features not only with their network neighbors but also with proteins that are active in similar tissues.…”

SeBioGraph: Semi-supervised Deep Learning for the Graph via Sustainable Knowledge Transfer

“…Biological graphs are notoriously complex and hard to decipher. Until now, many biomedical graph analytic methods have been proposed to analyze it (Grover and Leskovec, 2016 ; Fan et al, 2018 ; Zhang et al, 2018b ). Most of these approaches transform the original data into vectorial data.…”