Modeling the functional relationship network at the splice isoform level through heterogeneous data integration

Abstract

Functional relationship networks, which reveal the collaborative roles between genes, have significantly accelerated our understanding of gene functions and phenotypic relevance. However, establishing such networks for alternatively spliced isoforms remains a difficult, unaddressed problem due to the lack of systematic functional annotations at the isoform level, which renders most supervised learning methods difficult to be applied to isoforms. Here we describe a novel multiple instance learning-based probabilistic approach that integrates large-scale, heterogeneous genomic datasets, including RNA-seq, exon array, protein docking and pseudo-amino acid composition, for modeling a global functional relationship network at the isoform level in the mouse. Using this approach, we formulate a gene pair as a set of isoform pairs of potentially different properties. Through simulation and cross-validation studies, we showed the superior accuracy of our algorithm in revealing the isoform-level functional relationships. The local networks reveal functional diversity of the isoforms of the same gene, as demonstrated by both large-scale analyses and experimental and literature evidence for the disparate functions revealed for the isoforms of Ptbp1 and Anxa6 by our network. Our work can assist the understanding of the diversity of functions achieved by alternative splicing of a limited set of genes in mammalian genomes, and may shift the current gene-centered network prediction paradigm to the isoform level. Author summaryProteins carry out their functions through interacting with each other. Such interactions can be achieved through direct physical interactions, genetic interactions, or co-regulation. To summarize these interactions, researches have established functional relationship networks, in which each gene is represented as a node and the connections between the nodes represent how likely two genes work in the same biological process. Currently, these networks are established at the gene level only, while each gene, in mammalian systems, can be alternatively spliced into multiple isoforms that may have drastically different interaction partners. This information can be mined through integrating data that provide isoform-level information, such as RNA-seq and protein docking scores predicted from amino acid sequences. In this study, we developed a novel algorithm to integrate such data for predicting isoform-level functional relationship networks, which allows us to investigate the collaborative roles between genes at a high resolution.

Revisiting the identification of canonical splice isoforms through integration of functional genomics and proteomics evidence

“…Within each dataset, transcripts with missing values occurring in more than 50% experiments were removed for ensuring accurate expression value estimation. Of the 41, 11 RNA‐seq datasets were used to build the functional relationship network at the splice isoform level for the mouse in our previous study . The remaining 30 datasets were used as an independent test set for analyzing the expression behaviors of HCIs and NCIs in this study (Supporting Information File 1 for the full description of the RNA‐seq datasets used, Supporting Information Fig.…”

“…Within each dataset, transcripts with missing values occurring in more than 50% experiments were removed for ensuring accurate expression value estimation. Of the 41, 11 RNA-seq datasets were used to build the functional relationship network at the splice isoform level for the mouse in our previous study [27].…”

“…At the functional relationship network level, the highest connected isoforms (HCIs) are identified based on cofunctional connections with many other genes in the same biological pathways or processes. This network was built by integrating heterogeneous functional genomic data from RNA‐seq, exon array, protein docking, and pseudoamino acid composition . We performed a genome‐scale analysis of the local isoform‐level networks for 3427 validated multi‐isoform mouse genes (based on RefSeq database v37.2), and identified the HCIs for each multi‐isoform gene.…”

A Network of Splice Isoforms for the Mouse

“…In this paper, we used a Bayesian network-based multiple-instance learning (MIL) algorithm to solve this problem. MIL formulates a gene pair as a bag of multiple isoform pairs of potentially different probabilities to be functionally related, in analogy to a gene considered as a bag of isoforms of different functions 15 30 31 32 . Our work is focused on constructing isoform-level functional relationship networks (FRN) for the mouse, which significantly differs from previous studies 11 12 in terms of both computational approaches and research content.…”

Annotation of Alternatively Spliced Proteins and Transcripts with Protein-Folding Algorithms and Isoform-Level Functional Networks

“…Further, Li et al constructed a genome-wide functional relationship network for the mouse [54, 55] with the following steps: Collect isoform-pair feature data: RNA-Seq, exon array, pseudo-amino acid composition (pseudoAAC), and protein-docking data. For RNA-Seq and exon array, isoform expression was estimated followed by calculating isoform correlation as feature data.…”