Machine Learning in Bioinformatics

DMLDA-LocLIFT: Identification of multi-label protein subcellular localization using DMLDA dimensionality reduction and LIFT classifier

The paper introduces DMLDA-LocLIFT, a method for predicting multi-label protein subcellular localization using dimensionality reduction and a classifier, achieving high accuracy on bacterial and plant datasets.

Self-Supervised Contrastive Learning of Protein Representations By Mutual Information Maximization

CPCProt pretrains protein embeddings by telling a protein's own next fragment from fragments of random proteins, reportedly matching benchmarked self-supervised models with 2-10% of their parameters.

DeeplyEssential: A Deep Neural Network for Predicting Essential Genes in Microbes

This paper introduces a deep neural network for predicting essential genes in microbes using only protein sequences, outperforming existing methods that rely on structural or topological features.

Improving protein function prediction with synthetic feature samples created by generative adversarial networks

This work introduces FFPred-GAN, a GAN-based method to generate synthetic protein feature samples, improving function prediction accuracy across all three Gene Ontology domains.

DELPHI: accurate deep ensemble model for protein interaction sites prediction

DELPHI is a deep learning model that improves the accuracy of predicting protein-protein interaction sites by using an ensemble structure and novel features.

AE-LGBM: Sequence-Based Novel Approach To Detect Interacting Protein Pairs via Ensemble of Autoencoder and LightGBM.

AE-LGBM predicts protein-protein interactions using an ensemble of autoencoder and LightGBM, achieving high accuracy across multiple organisms and PPI networks.

Protocol for clustering of non-unified protein sequences through memory-map guided deep learning

The paper establishes a protocol for clustering non-unified protein sequences using memory-map guided deep learning, validated on human hormonal system disease-related genes.

Unsupervised protein embeddings outperform hand-crafted sequence and structure features at predicting molecular function

The paper shows that unsupervised protein sequence embeddings outperform traditional features in predicting molecular function, with a simple model achieving state-of-the-art results.

A Deep Learning Framework for Predicting Human Essential Genes by Integrating Sequence and Functional data

This paper introduces DeepSF, a deep learning framework that predicts human essential genes by integrating sequence and functional data, achieving high accuracy and outperforming traditional methods.

Prediction of the receptorome for the human-infecting virome

The paper predicts the receptorome for the human-infecting virome using a random-forest model and protein-protein interaction data, identifying 1380 human cell membrane proteins and 693 virus-specific receptors.

Alignment-free machine learning approaches for the lethality prediction of potential novel human-adapted coronavirus using genomic nucleotide

This study develops alignment-free machine learning models to predict the lethality of novel human coronaviruses using genomic nucleotide sequences, achieving 96.7% accuracy on existing strains.

Putative linear motifs mediate the trafficking to apical and basolateral membranes

The paper identifies linear motifs that mediate protein trafficking to apical and basolateral membranes using machine learning models trained on human transmembrane proteins.

Crowdsourcing Machine Intelligence Solutions to Accelerate Biomedical Science: Lessons learned from a machine intelligence ideation contest to improve the prediction of 3D domain swapping

The paper describes a machine intelligence competition aimed at improving 3D domain swapping prediction, resulting in enhanced predictive modeling through crowdsourced solutions.

Improving enzyme optimum temperature prediction with resampling strategies and ensemble learning

This study improves enzyme optimum temperature prediction by applying ensemble learning and resampling strategies to address data imbalance, reducing error on high Topt values and increasing overall R2.

An in silico approach to identification, categorization and prediction of nucleic acid binding proteins

The paper builds a database of nucleic acid binding proteins and predicts their functions using machine learning with 77.43% accuracy.

BERTology Meets Biology: Interpreting Attention in Protein Language Models

One attention head of a protein BERT model tracks 3D contacts (28% vs 1.3% background); other heads target binding sites, and deeper layers attend to higher-level properties.

iPromoter-BnCNN: a novel branched CNN based predictor for identifying and classifying sigma promoters

The paper introduces iPromoter-BnCNN, a branched CNN model for identifying and classifying six types of sigma promoters, showing superior performance on benchmark and independent test datasets.

CoRNeA: A pipeline to decrypt the protein-protein interaction from amino acid sequence information

CoRNeA predicts protein-protein interaction interfaces from amino acid sequences using co-evolution, machine learning, and network analysis, achieving higher precision and recall in eukaryotic complexes.

Assessing predictors for new post translational modification sites: a case study on hydroxylation

The paper evaluates hydroxylation site predictors using independent datasets, finding they perform no better than random, highlighting issues with self-reported performance and generalization.

AntiCP 2.0: An updated model for predicting anticancer peptides

AntiCP 2.0 predicts anticancer peptides using machine learning models trained on two datasets, achieving high MCC and AUROC scores on training data.

Transforming the Language of Life: Transformer Neural Networks for Protein Prediction Tasks

A RoBERTa-style Transformer, pre-trained on 450K Swiss-Prot sequences in 18 hours on 4 V100 GPUs, rivals DeepFam at family classification and beats PIPR at human protein-protein interaction prediction.

Prediction of protein subcellular localization using deep learning and data augmentation

This study improves protein subcellular localization prediction by replacing SVM with BiLSTM and SeqGAN, achieving 75% precision on a dataset of 8230 protein sequences.

iterb-PPse: Identification of transcriptional terminators in bacterial by incorporating nucleotide properties into PseKNC

The paper introduces iterb-PPse, a method for identifying bacterial transcriptional terminators using nucleotide properties and machine learning, achieving high accuracy on benchmark datasets.

Photosynthetic protein classification using genome neighborhood-based machine learning feature

The paper develops a machine learning method using genome neighborhood features to classify photosynthetic proteins, achieving 94% accuracy and outperforming existing tools.

ProtTox: Toxin identification from Protein Sequences

ProtTox is a deep learning method for toxin classification of protein sequences, achieving an F1-score 5% higher than existing baselines.

Learning Context-aware Structural Representations to Predict Antigen and Antibody Binding Interfaces

The paper develops a deep learning framework to predict antibody-antigen binding interfaces using graph convolutions, attention layers, and transfer learning, achieving state-of-the-art performance.

SAPH-ire TFx - A Recommendation-based Machine Learning Model Captures a Broad Feature Landscape Underlying Functional Post-Translational Modifications

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

Language modelling for biological sequences - curated datasets and baselines

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.

Mutation effect estimation on protein-protein interactions using deep contextualized representation learning

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.

iHyd-ProSite: A novel Computational Approach for Identifying Hydroxylation Sites in Proline Via Mathematical Modeling

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