Machine Learning in Bioinformatics

SignalP 6.0 predicts all five types of signal peptides with improved accuracy and applies to metagenomic data.

ProteinPrompt predicts protein-protein interactions using machine learning, offering two methods and consensus predictions for reliable results.

The paper introduces cross-attention PHV, a computational predictor for human-virus protein-protein interactions, which outperforms existing models and accurately predicts interactions for unknown viruses.

The paper develops function-conditioned generative models to predict functionally important protein residues using Gene Ontology data and protein language models.

This paper predicts protein-protein interactions using sequence encodings, achieving performance comparable to existing methods without complex physico-chemical features.

fastMSA accelerates multiple sequence alignment by using a dual encoder based on a protein language model, achieving a 34-fold speed-up while maintaining homolog recall and compatibility with downstream tasks like AlphaFold.

DSResSol is a deep learning model that predicts protein solubility from sequence, outperforming existing methods by 5-13% in accuracy and identifying key amino acids contributing to solubility.

EGRET improves PPI site prediction using a graph attention network and transfer learning from pre-trained models.

EnZymClass predicts plant acyl-ACP thioesterase substrate specificity using ensemble learning, achieving high accuracy and outperforming existing methods.

SBP-SITA is a sequence-based tool for predicting S-itaconation sites, achieving an AUC of 0.73 using an ensemble classifier.

The paper introduces bindEmbed21, a deep learning method using protein embeddings to predict binding residues for metal ions, nucleic acids, and small molecules, achieving high accuracy without using MSAs or structural data.

The paper introduces GraphSite, a method using AlphaFold2-predicted structures and graph transformers to predict DNA-binding residues, outperforming existing sequence and structure-based methods.

DeepANIS predicts antibody paratopes by integrating BiLSTM and transformer networks on concatenated CDR sequences, outperforming existing methods.

ACP-MHCNN is a deep convolutional neural network for anticancer peptide identification that outperforms existing methods in multiple metrics.

The paper introduces an unsupervised method using Transformer models, UMAP, and spectral clustering to organize protein sequences into biologically meaningful clusters.

The paper shows that a single attention layer can predict protein contacts as well as Potts models and pretrained Transformers, suggesting attention models can capture meaningful protein features from unlabeled data.

The paper develops a multitask transfer learning framework to predict virus-human protein interactions using statistical protein representations from a large sequence corpus, showing improved performance on benchmark datasets.

The paper develops an ensemble machine learning method to predict membrane-penetrating amino acids in peripheral membrane proteins, achieving high accuracy on an independent test set.

NetSolP predicts protein solubility in E. coli from sequence using transformers, achieving state-of-the-art performance with reduced dataset bias.

AutoGenome is an automated deep learning framework for genomic profiling data that uses a residual fully-connected neural network and AutoML to train high-performance models.

The paper introduces Multi-Branch-CNN, a parallel CNN method for classifying ion channel interacting peptides, showing improved performance on novel sequences compared to traditional ML algorithms and other CNN variants.

The paper develops machine learning models using sequence-derived features to predict protein abundance across multiple cell lines and identifies key features for this task.

This paper introduces GAT-GO, a graph attention network that improves protein function prediction using predicted structure information, outperforming existing methods on benchmark datasets.

USPNet is a signal peptide predictor using a deep protein language model that outperforms existing methods, especially in metagenomic data without additional group information.

The paper develops a machine learning model to predict recombinant protein expression in Bacillus subtilis using deep protein representations, achieving moderate performance for prioritizing expression candidates.

This paper investigates deep feedforward neural networks for classifying transposon-derived piRNAs, showing improved performance over existing methods with an F-measure of 0.872.

This study introduces LM-GVP, a deep learning framework combining a protein Language Model and Graph Neural Network to predict protein properties from sequence and structure, outperforming existing methods on fluorescence, protease stability, and Gene Ontology function prediction.

PARROT is a deep learning framework for analyzing large protein datasets, capable of classification and regression tasks using only raw protein sequences.

IAV-CNN predicts influenza A virus antigenic variants using a 2D convolutional neural network with a new amino acid representation, achieving state-of-the-art performance on three datasets.

TripletProt improves protein functional annotation using Siamese networks, achieving better F1 scores than existing methods in sub-cellular localization and gene ontology prediction.