FINER: enhancing the prediction of tissue-specific functions of isoforms by refining isoform interaction networks

Abstract

Annotating the functions of gene products is a mainstay in biology. A variety of databases have been established to record functional knowledge at the gene level. However, functional annotations at the isoform resolution are in great demand in many biological applications. Although critical information in biological processes such as protein–protein interactions (PPIs) is often used to study gene functions, it does not directly help differentiate the functions of isoforms, as the ‘proteins’ in the existing PPIs generally refer to ‘genes’. On the other hand, the prediction of isoform functions and prediction of isoform–isoform interactions, though inherently intertwined, have so far been treated as independent computational problems in the literature. Here, we present FINER, a unified framework to jointly predict isoform functions and refine PPIs from the gene level to the isoform level, enabling both tasks to benefit from each other. Extensive computational experiments on human tissue-specific data demonstrate that FINER is able to gain at least 5.16% in AUC and 15.1% in AUPRC for functional prediction across multiple tissues by refining noisy PPIs, resulting in significant improvement over the state-of-the-art methods. Some in-depth analyses reveal consistency between FINER’s predictions and the tissue specificity as well as subcellular localization of isoforms.

IsoFrog: a reversible jump Markov Chain Monte Carlo feature selection-based method for predicting isoform functions

“…Dataset A ( Chen et al 2019 ) contains expression data derived from 4643 human RNA-seq experiments from the NCBI Sequence Read Archive (SRA) ( Leinonen et al 2011 ). The expressions are measured in Transcripts Per Million.…”

“…RNA-seq data, for isoform function prediction are in urgent demand. Due to the lack of transcription-level annotations, many semi-supervised learning methods have been applied in isoform function prediction, such as mi-SVM ( Eksi et al 2013 ), iMILP ( Li et al 2014 ), WLRM ( Luo et al 2017 ), DeepIsoFun ( Shaw et al 2019 ), DIFFUSE ( Chen et al 2019 ), IsoFun ( Yu et al 2020 ), DisoFun ( Wang et al 2020 ), and IsoResolve ( Li et al 2021 ). These methods use the expression profiles as the primary input and Gene Ontology (GO) ( Harris et al 2004 ) terms as the function labels, and can be divided into four categories: multiple instance learning (MIL)-based methods, network propagation-based methods, matrix factorization-based methods, and domain adaptation (DA)-based methods.…”

An algorithmic framework for isoform-specific functional analysis

“…We used the prediction of GO terms for human that are provided with the tool. We also tried to use the predictions provided by FINER [40], however none of them matched entries in our collection.…”