Machine Learning in Bioinformatics

The paper introduces SAPH-ire TFx, a machine learning model that recommends functional post-translational modifications by analyzing multiple features.

The paper analyzes properties of protein datasets for language modelling and evaluates n-gram and RNN models, finding that organism domain and data quality affect performance while completeness has little impact.

This paper introduces MuPIPR, a deep learning framework that estimates mutation effects on protein-protein interactions by using contextualized amino acid representations and a Siamese residual recurrent convolutional encoder.

This paper introduces a computational predictor for identifying hydroxylation sites in proline residues using mathematical modeling and validates its accuracy through multiple tests.

The paper introduces MTDsite, a multi-task deep learning method for predicting protein-binding residues across DNA, RNA, peptide, and carbohydrate, achieving higher accuracy than existing methods.

The paper introduces a Transformer-based model for annotating Gene Ontology terms in protein sequences, showing improved accuracy over CNN-based methods and incorporating additional features like BLAST motifs and structural data.

SAMbinder predicts SAM binding residues in proteins from amino acid sequences using machine learning, achieving MCC 0.61 and AUROC 0.89 on a realistic dataset.

This study develops a hybrid model combining BLAST and machine learning to predict pattern recognition receptors, achieving high accuracy with an MCC of 0.82 and AUROC of 0.95.

This paper introduces an end-to-end multitask learning approach using language models to predict protein structure and function without evolutionary alignments, achieving faster predictions with moderate accuracy.

TranCEP predicts transmembrane transport protein substrates by integrating amino acid composition, evolutionary information, and positional data, outperforming existing methods.

DeepKinZero predicts kinase-phosphosite associations for understudied kinases using zero-shot learning, improving accuracy compared to existing methods.

UDSMProt uses self-supervised language modeling to create a universal protein sequence model that outperforms task-specific methods in enzyme class prediction, gene ontology prediction, and remote homology detection.

This paper introduces two deep learning methods, DeepDBP-ANN and DeepDBP-CNN, to identify DNA-binding proteins from protein sequences, achieving state-of-the-art accuracy on benchmark datasets.

This paper introduces ECNet, a deep learning model that uses evolutionary context to predict protein variant fitness, outperforming existing methods on multiple datasets.

The paper introduces SMA, an algorithm to predict G domains in G proteins by matching G-box patterns and inter-box spacer lengths, improving precision for family-specific motifs.

The paper develops machine learning models to predict antibody biophysical properties using sequence data, finding that charge and hydrophobicity are key features.

This paper introduces SeqVec, a method to represent protein sequences as continuous vectors using deep learning, achieving better prediction results than methods relying on evolutionary information.

FastFeatGen improves DNA N6-methyladenine site prediction by combining parallel feature extraction and machine learning, achieving high accuracy on rice and mouse genomes.

The paper compares predicted protein localization spectra across ten eukaryotic species to infer evolutionary relationships.

The paper develops a machine learning tool, PSPredictor, to predict phase separation proteins using sequence information, achieving high accuracy and outperforming existing methods.

The paper develops an artificial neural network model to predict breast cancer-related proteins, achieving high accuracy and identifying top-ranked proteins in three categories.

PerMemDB is a database of eukaryotic peripheral membrane proteins, containing 241,173 entries from 1,216 organisms, annotated with interaction data and domain predictions using MBPpred.

This paper introduces SKIPHOS, a non-kinase specific phosphorylation site predictor using random forests and amino acid skip-gram embeddings, showing better performance than existing methods for three phosphorylation residues.

This paper introduces expectation pooling, an interpretable CNN pooling method that improves DNA-protein binding prediction by combining probabilistic ideas with deep learning.

The paper introduces TargetP 2.0, a deep learning method to identify N-terminal sorting signals in proteins, revealing insights into the importance of the second residue in targeting peptides.

The paper improves protein thermal stability classification using sequence mining and network centrality statistics, achieving 96% accuracy.

This study introduces node2loc, a deep learning method that predicts protein subcellular locations using protein-protein network representations, achieving an MCC of 0.812.

DeepGOPlus improves protein function prediction from sequence alone, achieving faster and more accurate results on biological processes and cellular components.

Across five protein tasks with evolution-aware splits, self-supervised pretraining helped nearly every model on every task, yet alignment-based features still beat learned embeddings on secondary structure and contact prediction.

The paper introduces an anti-cancer peptide prediction tool, ACS, achieving 95% accuracy, outperforming existing methods.