Functional protein representations from biological networks enable diverse cross-species inference

Abstract

Transferring knowledge between species is key for many biological applications, but is complicated by divergent and convergent evolution. Many current approaches for this problem leverage sequence and interaction network data to transfer knowledge across species, exemplified by network alignment methods. While these techniques do well, they are limited in scope, creating metrics to address one specific problem or task. We take a different approach by creating an environment where multiple knowledge transfer tasks can be performed using the same protein representations. Specifically, our kernel-based method, MUNK, integrates sequence and network structure to create functional protein representations , embedding proteins from different species in the same vector space. First we show proteins in different species that are close in MUNK-space are functionally similar. Next, we use these representations to share knowledge of synthetic lethal interactions between species . Importantly, we find that the results using MUNK-representations are at least as accurate as existing algorithms for these tasks. Finally, we generalize the notion of a phenolog (‘orthologous phenotype’) to use functionally similar proteins (i.e. those with similar representations). We demonstrate the utility of this broadened notion by using it to identify known phenologs and novel non-obvious ones supported by current research.

Data-driven biological network alignment that uses topological, sequence, and functional information

“…MUNK also links the original networks via anchors, but it uses matrix factorization to obtain an alignment [ 36 ]. In our preliminary analyses, MUNK’s similarity scores could not distinguish between functionally related and functionally unrelated proteins.…”

NetQuilt: Deep Multispecies Network-based Protein Function Prediction using Homology-informed Network Similarity

“…Other features that can be constructed over nodes include graphlets [18], and random walk profiles of nodes within their graph, which have been extended and applied to heterogeneous and multiplex biological networks [19,20]. Network embedding has been extensively used in protein functional analysis and includes methods based on matrix factorization [4], graph kernels [21] and deep learning [12,22,23]. A comprehensive review of network embedding in computational biology compared to other types of network-based algorithms for several applications can be found in [24], and reviews of network representation learning methods in general can be found in [25] and [26].…”

“…Ideally, a protein function prediction method should be able to use homology information to supplement network information even on proteins whose sequences are not similar to the training set protein sequences. Another method, MUNK, is a kernel-based method that produces functional embeddings used for predicting synthetic lethality for pairs of proteins of multiple species [21]; they additionally demonstrate that proteins close in this embedding space are similar in function. The key idea of their approach is that proteins from different species are embedded in the same vector space using graph kernels with landmark proteins in the networks of the two species that perform the same functions.…”

Advances in synthetic lethality for cancer therapy: cellular mechanism and clinical translation

“…The predictions are produced by scoring gene pair interaction by using expression profiles data, screening results for shRNA, and copy number data [83]. Since the human genome-wide evaluation of synthetic lethal interactions is still impractical, computational data transfer of model organism interactions, ranging from yeast to humans, became the study trend [84,85]. New innovative programs and computational methods are now being developed to enable the rapid analysis of data sets for gene expression.…”