Machine Learning in Bioinformatics

ProtFun: A Protein Function Prediction Model Using Graph Attention Networks with a Protein Large Language Model

ProtFun predicts protein functions using a graph attention network with protein LLM embeddings, outperforming existing methods on benchmark datasets.

DDGemb: predicting protein stability change upon single- and multi-point variations with embeddings and deep learning

DDGemb predicts protein stability changes from single- and multi-point variations using embeddings and deep learning, achieving state-of-the-art performance.

KSMoFinder - Knowledge graph embedding of proteins and motifs for predicting kinases of human phosphosites

KSMoFinder predicts kinase-phosphosite relationships by integrating protein context and motif composition using a knowledge graph embedding approach, outperforming existing methods.

How suitable are clustering methods for functional annotation of proteins?

This study evaluates nine clustering algorithms for unsupervised protein function annotation using pseudo-amino acid composition features, finding some methods effective for functional family segregation but less so for fine-grained differences.

Transcription factor prediction using protein 3Dsecondary structures

The paper introduces a deep learning method, StrucTFactor, that uses 3D protein structures to predict transcription factors, outperforming existing methods by 17% in Matthews correlation coefficient.

StackGlyEmbed: Prediction of N-linked Glycosylation sites using protein language models

StackGlyEmbed predicts N-linked glycosylation sites using a stacking ensemble model, achieving high performance metrics and outperforming existing methods.

Aggregating Residue-Level Protein Language Model Embeddings with Optimal Transport

This paper introduces an optimal transport-based method to aggregate residue-level protein language model embeddings into fixed-length representations, showing improved performance over average pooling for downstream tasks.

MULAN: Multimodal Protein Language Model for Sequence and Structure Encoding

The paper introduces MULAN, a multimodal protein language model that integrates sequence and structure information, showing improved performance on downstream tasks compared to sequence-only and structure-aware models.

PERSEUcpp: A machine learning strategy to predictcell-penetrating peptides and their uptake efficiency

PERSEUcpp predicts cell-penetrating peptides and their uptake efficiency using machine learning, achieving high accuracy and interpretability.

GENA-LM: A Family of Open-Source Foundational DNA Language Models for Long Sequences

GENA-LM introduces transformer-based DNA language models for long sequences, capable of handling up to 36,000 base pairs and supporting multispecies and taxon-specific applications.

Leveraging ancestral sequence reconstruction for protein representation learning

This paper introduces LASE, a method using ancestral sequence reconstruction to train protein language models with higher accuracy and faster computation than large models.

Tokenization to Transfer: Do Genomic Foundation Models Learn Good Representations?

This study evaluates seven Genomic Foundation Models across 52 genomic tasks, finding that random baselines are strong and tokenizer choices significantly impact performance.

OPUS-GO: An interpretable protein/RNA sequence annotation framework based on biological language model

OPUS-GO provides sequence-level and residue-level annotations for proteins and RNA sequences using a biological language model and modified MIL strategy, improving accuracy and interpretability.

EnzymeCAGE: A Geometric Foundation Model for Enzyme Retrieval with Evolutionary Insights

EnzymeCAGE is a geometric foundation model trained on enzyme-reaction pairs to predict enzyme function, de-orphan reactions, identify catalytic sites, and reconstruct biosynthetic pathways.

PepMSND: Integrating Multi-level Feature Engineering and Comprehensive Databases to Enhance in vivo/in vitro Peptide Blood Stability Prediction

The paper develops a deep learning model called PepMSND to predict peptide blood stability using multi-level feature engineering and a comprehensive database of 635 samples, achieving high accuracy and AUC.

S-PLM: Structure-aware Protein Language Model via Contrastive Learning between Sequence and Structure

S-PLM improves protein language models by integrating 3D structure information through contrastive learning, outperforming sequence-only models in clustering and classification tasks.

Accurate nucleic acid-binding residue identification based on domain-adaptive protein language model and explainable geometric deep learning

GeSite identifies nucleic acid-binding residues using a domain-adaptive protein language model and explainable geometric deep learning, showing superior performance on benchmark datasets.

L2G: Repurposing Language Models for Genomics Tasks

This paper introduces L2G, a method that repurposes pre-trained language models for genomics tasks, achieving better performance than specialized models on multiple benchmarks.

Evaluating generalizability of artificial intelligence models for molecular datasets

This paper evaluates how well deep learning models generalize to new molecular sequences using a new framework called SPECTRA across 18 datasets and 19 models.

Employing Machine Learning Techniques to Detect Protein-Protein Interaction: A Survey, Experimental, and Comparative Evaluations

This paper surveys machine learning techniques for protein-protein interaction detection, introduces a hierarchical taxonomy, and evaluates algorithms through empirical and experimental assessments.

Protein2Text: Providing Rich Descriptions for Protein Sequences

The paper introduces BetaDescribe, a model that generates rich textual descriptions of proteins, enhancing understanding of their functions and properties.

Accurate and General DNA Representations Emerge from Genome Foundation Models at Scale

This paper introduces AIDO.DNA, a DNA language model that improves representation of DNA sequences across multiple tasks by scaling model size and training data.

The Nucleotide Transformer: Building and Evaluating Robust Foundation Models for Human Genomics

The paper introduces the Nucleotide Transformer, a foundation model for genomics, showing it can predict molecular phenotypes from DNA sequences with transfer learning across tasks and organisms.

Limitations and Enhancements in Genomic Language Models: Dynamic Selection Approach

This paper proposes a dynamic model selection approach for genomic language models, showing improved performance through multi-model fusion and identifying limitations related to motif reliance.

Prediction of plant resistance proteins using alignment-based and alignment-free approaches

The paper develops a hybrid model combining alignment-based and alignment-free methods to predict plant resistance proteins, achieving high AUROC scores and creating an online platform for this purpose.

Harnessing Contextual Embeddings: A Deep Learning Framework for Predicting PCR Amplification Using BERT Tokenization

This paper predicts PCR amplification outcomes using a deep learning framework combining BERT tokenization and a CNN-BiLSTM model, achieving 84.8% accuracy and improved sensitivity, specificity, and AUC.

Benchmarking Recent Computational Tools for DNA-binding Protein Identification

This paper benchmarks eleven computational tools for DNA-binding protein identification, highlights limitations, and recommends improvements, providing a consensus method and datasets.

Prediction of virus-host association using protein language models and multiple instance learning

This paper introduces EvoMIL, a deep learning method that predicts virus-host associations using protein language models and multiple instance learning, achieving improved accuracy in both prokaryotic and eukaryotic hosts.

Ranking protein-peptide binding affinities with protein language models

This study uses protein language models to rank protein-peptide binding affinities without labeled data or structural information, showing agreement with experimental IC50 values across five datasets.

Enhanced prediction of protein functional identity through the integration of sequence and structural features

The paper develops a method to predict if two proteins catalyze the same reaction by integrating sequence and structural features, outperforming existing approaches.