Predicting protein subcellular location using learned distributed representations from a protein-protein network

Abstract

Functions of proteins are in general related to their subcellular locations. To identify the functions of a protein, we first need know where this protein is located. Interacting proteins tend to locate in the same subcellular location. Thus, it is imperative to take the protein-protein interactions into account for computational identification of protein subcellular locations.In this study, we present a deep learning-based method, node2loc, to predict protein subcellular location. node2loc first learns distributed representations of proteins in a protein-protein network using node2vec, which acquires representations from unlabeled data for downstream tasks. Then the learned representations are further fed into a recurrent neural network (RNN) to predict subcellular locations. Considering the severe class imbalance of different subcellular locations, Synthetic Minority Over-sampling Technique (SMOTE) is applied to artificially boost subcellular locations with few proteins.We construct a benchmark dataset with 16 subcellular locations and evaluate node2loc on this dataset. node2loc yields a Matthews correlation coefficient (MCC) value of 0.812, which outperforms other baseline methods. The results demonstrate that the learned presentations from a proteinprotein network have strong discriminate ability for classifying protein subcellular locations and the RNN is a more powerful classifier than traditional machine learning models. node2loc is freely available at https://github.com/xypan1232/node2loc.

ProtFinder: finding subcellular locations of proteins using protein interaction networks

“…The choice of LSTM cell in each binary classifier is motivated by Pan et. al [5]. However, it is not justified.…”

“…Motivated by Pan et. al [5], binary classifier BC i consists of an LSTM cell [28] which learns the hidden patterns within the generated protein embeddings. This is followed by a fully-connected layer which creates non-linear combinations of the learnt hidden patterns.…”

“…While these pipelines rely on experimental validation, they can be used to add scalability to the robustness of experimentation. Some of the earlier computational methods treated this task as a multi-class single-label classification [3,4,5]. However, with recent advancements it was learnt that a protein can either co-exist in multiple locations within a cell or move around different subcellular components [6].…”

“…In 2019, Pan et. al [5] developed a deep learning (RNN) single-label classifier called node2loc and used the protein interaction networks to classify across 16 subcellular locations.…”

lncLocPred: Predicting LncRNA Subcellular Localization Using Multiple Sequence Feature Information

“…LncSLdb has collected lncRNA subcellular localization information for more than 11,000 transcripts from three species. Although methods to predict the subcellular localization of lncRNAs are limited compared with that of proteins [25]- [29], some achievements have been made so far. The existing methods to identify subcellular localization fall into two categories, biochemical experiments and computational methods.…”