Machine Learning in Bioinformatics

Limitations of Current Machine-Learning Models in Predicting Enzymatic Functions for Uncharacterized Proteins

The paper evaluates machine-learning models' ability to predict enzymatic functions for uncharacterized proteins and finds they fail to integrate scientific reasoning and make logic errors.

From Scarce Functional Labels to Label-Aware Generation in Homologous Protein Families

The paper develops a two-stage strategy for functional annotation and label-aware sequence generation in protein families using limited labels and phylogeny-aware evaluation.

NTxPred2: A large language model for predicting neurotoxic peptides and neurotoxins

The paper develops separate models for predicting neurotoxic peptides and neurotoxins using machine learning and protein language models, achieving high AUC scores and outperforming existing methods.

SuperMetal: A Generative AI Framework for Rapid and Precise Metal Ion Location Prediction in Proteins

SuperMetal predicts metal-binding sites in proteins with high precision and efficiency, outperforming existing models for zinc ions.

Plastic-hydrolytic enzyme classification using explainable deep learning

This paper introduces PEPIC, an explainable deep learning model for classifying plastic-hydrolytic enzymes from protein sequences, showing improved performance and biological validation.

EZpred: improving deep learning-based enzyme function prediction using unlabeled sequence homologs

EZpred improves enzyme function prediction by using unlabeled sequence homologs, achieving higher F1-scores than existing methods.

EPEPDI: prediction of binding free energy changes from missense mutations in double and single-stranded DNA-binding proteins

EPEPDI predicts binding free energy changes from missense mutations in DNA-binding proteins by distinguishing double and single-stranded DNA interactions and using deep learning with sequence embeddings.

Proteins Need Extra Attention: Improving the Predictive Power of Protein Language Models on Mutational Datasets with Hint Tokens

This paper introduces hint token learning to improve protein language models on mutational datasets through three case studies, showing better performance than traditional fine-tuning.

Mechanism-Aware Inductive Bias Enhances Generalization in Protein-Protein Interaction Prediction

PLMDA-PPI improves protein-protein interaction prediction by integrating biophysical inductive biases, outperforming existing models in efficiency and generalization.

Machine learning-based structural classification of lytic polysaccharide monooxygenases

The paper trains a machine learning model to classify LPMO sequences into CAZy families using structural features extracted from 3D models.

Medium-sized protein language models perform well at transfer learning on realistic datasets

Medium-sized protein language models (e.g. ESM C 600M) with mean embeddings nearly match ESM-2 15B and ESM C 6B in transfer learning on mutational scanning and PISCES data.

MKFGO: Integrating Multi-Source Knowledge Fusion with Pre-Trained Language Model for High-Accuracy Protein Function Prediction

MKFGO integrates multi-source biological data with deep learning to predict protein functions, outperforming existing methods on 1522 proteins.

NEAR: Neural Embeddings for Amino acid Relationships

NEAR improves homology detection speed and accuracy by using neural embeddings trained with contrastive learning, outperforming existing methods as a pre-filter for protein sequence annotation.

MutBERT: Probabilistic Genome Representation Improves Genomics Foundation Models

MutBERT improves genomic foundation models by using a probabilistic representation of SNPs, outperforming existing models on downstream tasks.

PLMFit : Benchmarking Transfer Learning with Protein Language Models for Protein Engineering

PLMFit benchmarks transfer learning methods on protein language models for protein engineering tasks, identifying optimal strategies for knowledge transfer.

Allosteric Site Prediction Using Protein Language Models and Orthosteric Conditioning

The paper predicts allosteric binding sites using protein language models with orthosteric conditioning, outperforming static embeddings and achieving results comparable to structure-based methods.

Genomic Touchstone: Benchmarking Genomic Language Models in the Context of the Central Dogma

Genomic Touchstone evaluates genomic language models across 36 tasks and 88 datasets, revealing insights on model performance, scaling, and pretraining strategies.

RC-GNN: A predictive model of enzyme-reaction pairs

RC-GNN predicts enzyme-reaction pairs using amino acid sequences and reaction stoichiometry, achieving high ROC-AUC scores on diverse datasets.

AllerStack: Predicting Allergenic Proteins with a Stacked Ensemble Approach

AllerStack predicts allergenic proteins using a stacked ensemble model with high accuracy and a public web server for real-time predictions.

AstraROLE2 & AstraSUIT2: Multi-Task Annotation Models for Functional Profiling of Proteins

The paper introduces two transformer-based models for multi-task protein function annotation, achieving high performance on various functional labels.

Generalized Biological Foundation Model with Unified Nucleic Acid and Protein Language

This paper introduces LucaOne, a foundation model trained on nucleic acid and protein sequences, demonstrating understanding of biological principles and performing well on related tasks.

Squidly: Enzyme Catalytic Residue Prediction Harnessing a Biology-Informed Contrastive Learning Framework

Squidly predicts catalytic residues using a biology-informed contrastive learning framework, outperforming existing methods and integrating with BLAST for efficient annotation.

HAVEN: Hierarchical Attention for Viral protEin-based host iNference

HAVEN predicts viral hosts from protein sequences, achieving high accuracy for both common and rare hosts while outperforming larger models.

GeneChat: A Multi-Modal Large Language Model for Gene Function Prediction

GeneChat predicts gene function from DNA sequences using a multi-modal large language model, outperforming GPT-4o on standard metrics.

Sequence-based deep learning model for annotating lytic polysaccharide monooxygenase families

The paper introduces a deep learning model for annotating LPMO families, achieving high accuracy and outperforming existing tools.

ProtMamba: a homology-aware but alignment-free protein state space model

ProtMamba is a homology-aware, alignment-free protein language model that outperforms state-of-the-art models in homolog-conditioned sequence generation.

GOBoost: Leveraging Long-Tail Gene Ontology Terms for Accurate Protein Function Prediction

GOBoost improves protein function prediction by addressing long-tail distribution of Gene Ontology Terms using an ensemble strategy and specialized loss functions.

Pool PaRTI: A PageRank-Based Pooling Method for Identifying Critical Residues and Enhancing Protein Sequence Representations

Pool PaRTI improves protein sequence representations by using PageRank on transformer attention matrices to identify critical residues, enhancing performance and interpretability across ML tasks.

Protein sequence classification using natural language processing techniques

This paper applies NLP techniques to classify protein sequences using various machine learning and deep learning models, finding that ensemble methods and transformers achieve high accuracy.

Designing of thermostable proteins with a desired melting temperature

This paper develops models to predict and design proteins with desired melting temperatures using machine learning and large language models, achieving high accuracy and providing tools for application.