In-Pero: Exploiting Deep Learning Embeddings of Protein Sequences to Predict the Localisation of Peroxisomal Proteins.

Marco Anteghini,Vitor Martins Dos Santos,Edoardo Saccenti

doi:10.3390/ijms22126409

Marco Anteghini, Vitor Martins Dos Santos + Show 1 more

Open Access

https://doi.org/10.3390/ijms22126409

Copy DOI

Abstract

Peroxisomes are ubiquitous membrane-bound organelles, and aberrant localisation of peroxisomal proteins contributes to the pathogenesis of several disorders. Many computational methods focus on assigning protein sequences to subcellular compartments, but there are no specific tools tailored for the sub-localisation (matrix vs. membrane) of peroxisome proteins. We present here In-Pero, a new method for predicting protein sub-peroxisomal cellular localisation. In-Pero combines standard machine learning approaches with recently proposed multi-dimensional deep-learning representations of the protein amino-acid sequence. It showed a classification accuracy above 0.9 in predicting peroxisomal matrix and membrane proteins. The method is trained and tested using a double cross-validation approach on a curated data set comprising 160 peroxisomal proteins with experimental evidence for sub-peroxisomal localisation. We further show that the proposed approach can be easily adapted (In-Mito) to the prediction of mitochondrial protein localisation obtaining performances for certain classes of proteins (matrix and inner-membrane) superior to existing tools.

Highlights

In eukaryotes, there are ten main subcellular localisations which can be further subdivided into intra-organellar compartments
This has led to the hypothesis that each protein has evolved to function optimally in a given subcellular compartment, and to the idea that the information encoded in the sequence can be used to predict the subcellular localisation
We compared four commonly used machine learning approaches (Logistic Regression, Partial Least Squares Discriminant analysis, Random Forest and Support Vector Machines) in combination with different protein sequence encodings and embeddings to select the best classification strategy to predict the sub-localisation of peroxisomal proteins

Summary

Introduction

There are ten main subcellular localisations which can be further subdivided into intra-organellar compartments (see Figure 1A) These organelles perform one or more, and often complementary, specific tasks in the cellular machinery. It has been observed that proteins from different organelles show signatures, in their amino acid composition, that associate with their subcellular localisation [4]. This has led to the hypothesis that each protein has evolved to function optimally in a given subcellular compartment, and to the idea that the information encoded in the sequence can be used to predict the subcellular localisation

Methods

Results

Conclusion