Tissue-specific mouse mRNA isoform networks

Abstract

Alternative Splicing produces multiple mRNA isoforms of genes which have important diverse roles such as regulation of gene expression, human heritable diseases, and response to environmental stresses. However, little has been done to assign functions at the mRNA isoform level. Functional networks, where the interactions are quantified by their probability of being involved in the same biological process are typically generated at the gene level. We use a diverse array of tissue-specific RNA-seq datasets and sequence information to train random forest models that predict the functional networks. Since there is no mRNA isoform-level gold standard, we use single isoform genes co-annotated to Gene Ontology biological process annotations, Kyoto Encyclopedia of Genes and Genomes pathways, BioCyc pathways and protein-protein interactions as functionally related (positive pair). To generate the non-functional pairs (negative pair), we use the Gene Ontology annotations tagged with “NOT” qualifier. We describe 17 Tissue-spEcific mrNa iSoform functIOnal Networks (TENSION) following a leave-one-tissue-out strategy in addition to an organism level reference functional network for mouse. We validate our predictions by comparing its performance with previous methods, randomized positive and negative class labels, updated Gene Ontology annotations, and by literature evidence. We demonstrate the ability of our networks to reveal tissue-specific functional differences of the isoforms of the same genes. All scripts and data from TENSION are available at: 10.25380/iastate.c.4275191.

Imbalance deep multi‐instance learning for predicting isoform–isoform interactions

“…At the same time, an isoform–isoform interactions database (IIIDB) is constructed for further understanding protein functions and biological pathways. By integrating multiple tissue‐specific RNA‐seq data sets and sequence information, Kandoi et al 31 employed a Random Forest model to learn tissue‐specific mRNA isoform functional networks (TENSION) following a leave‐one‐tissue‐out strategy.…”

“…Apparently, studies 16,31 on single‐isoform interactions ignore the genes with multiple isoforms, which occupy a very large proportion of human transcriptome. DIIP 32 relies on domain annotations on known reference PPIs and isoforms to infer the isoform interactions.…”

“…Although the MIL‐based solutions 15,17 expand the isoform interactions prediction from single‐isoform to genome‐wide, they equally treat all isoform pairs of a gene bag, which is difficult to distinguish the interacting isoform pairs from noninteracting ones of a positive gene bag. In addition, these aforementioned solutions 16,31,15,17 implicitly target for a balanced III data set and ignore the intrinsic imbalanced problem. To address these issues, we proposed IDMIL‐III for predicting isoform interactions from genome‐wide.…”

“…In addition, we compare IDMIL‐III against IIIDB, 16 TENSION, 31 SIB‐MIL, 15 and DMIL‐III 17 . All the parameters are kept the same as the authors reported or optimized through grid search.…”

“…In practice, we focus more on the interacting isoform pairs than the noninteracting ones. However, existing solutions simply under‐sample the negative bags to keep data balanace and then induce classifier on the manually balanced data to predict IIIs, 15,16,31,17 whose prediction performance drops sharply when dealing with intrinsically imbalanced MIL data. In addition, the key instance (interacting isoform pair) plays an important role in triggering the label of the positive (interacting) gene bags, which is ignored by the current solutions.…”