mRNALoc: a novel machine-learning based in-silico tool to predict mRNA subcellular localization

Abstract

Recent evidences suggest that the localization of mRNAs near the subcellular compartment of the translated proteins is a more robust cellular tool, which optimizes protein expression, post-transcriptionally. Retention of mRNA in the nucleus can regulate the amount of protein translated from each mRNA, thus allowing a tight temporal regulation of translation or buffering of protein levels from bursty transcription. Besides, mRNA localization performs a variety of additional roles like long-distance signaling, facilitating assembly of protein complexes and coordination of developmental processes. Here, we describe a novel machine-learning based tool, mRNALoc, to predict five sub-cellular locations of eukaryotic mRNAs using cDNA/mRNA sequences. During five fold cross-validations, the maximum overall accuracy was 65.19, 75.36, 67.10, 99.70 and 73.59% for the extracellular region, endoplasmic reticulum, cytoplasm, mitochondria, and nucleus, respectively. Assessment on independent datasets revealed the prediction accuracies of 58.10, 69.23, 64.55, 96.88 and 69.35% for extracellular region, endoplasmic reticulum, cytoplasm, mitochondria, and nucleus, respectively. The corresponding values of AUC were 0.76, 0.75, 0.70, 0.98 and 0.74 for the extracellular region, endoplasmic reticulum, cytoplasm, mitochondria, and nucleus, respectively. The mRNALoc standalone software and web-server are freely available for academic use under GNU GPL at http://proteininformatics.org/mkumar/mrnaloc.

DM3Loc: multi-label mRNA subcellular localization prediction and analysis based on multi-head self-attention mechanism

“…( 19 ) proposed a machine-learning-based tool, named iLoc-mRNA that focused on the mRNA subcellular localization prediction of Homo sapiens, where a support vector machine (SVM) was applied to a combination of optimally preselected features. Recently, mRNALoc was developed to provide predictions for five subcellular locations of eukaryotic mRNAs using SVM on the pseudo K-tuple nucleotide composition features ( 20 ). In the real transcriptome world, mRNAs are localized in multiple compartments, as shown by data in an RNA subcellular localization database ( 21 ), in a way similar to that of about half of the proteins each localized in multiple compartments ( 22 ).…”

mLoc-mRNA: predicting multiple sub-cellular localization of mRNAs using random forest algorithm coupled with feature selection via elastic net

“…There are three tools such as RNATracker [ 34 ], iLoc-mRNA [ 35 ] and mRNALoc [ 36 ] available for predicting mRNA localizations. Since the RNATracker has been trained with the CeFra-Seq/APEX-RIP dataset involving gene expression and coordinate files, it was not included for comparison.…”

“…This may be the reason the existing tools are based on the supervised learning model. The supervised computational tools such as RNATracker [ 34 ], iLoc-mRNA [ 35 ], and mRNALoc [ 36 ] have been developed for predicting the mRNA localization. It is a well-known fact that a single mRNA could be present in more than one localization, whereas the existing tools are meant for predicting single localization only.…”

iT4SE-EP: Accurate Identification of Bacterial Type IV Secreted Effectors by Exploring Evolutionary Features from Two PSI-BLAST Profiles

“…As one of the most powerful supervised learning algorithms, SVM has been successfully applied to an increasingly wide variety of bioinformatics applications, especially the protein classification tasks [ 28 , 29 , 30 ]. Theoretically, SVM maps the training examples to points in a high-dimensional space so as to find the maximum-margin hyperplane which might separate the two categories.…”