Machine Learning in Bioinformatics

A machine learning model for the proteome-wide prediction of lipid-interacting proteins

This paper introduces SLiPP, a machine learning model that predicts lipid-binding pockets in proteins with high accuracy, identifying novel lipid-binding proteins and potential therapeutic targets.

Efficient Inference, Training, and Fine-tuning of Protein Language Models

This paper improves the computational efficiency of protein language models, achieving faster inference, reduced memory usage, and efficient fine-tuning for better accessibility.

Data-optimal scaling of paired antibody language models

This study derives a scaling law for antibody language models, showing that performance improvements require both large model size and sufficient data.

DeepSEA: an alignment-free deep learning tool for functional annotation of antimicrobial resistance proteins

This paper introduces DeepSEA, a deep learning tool that functionally annotates antimicrobial resistance proteins without sequence alignment, achieving higher recall than alignment-based methods.

Rapid prediction of thermodynamically destabilizing tyrosine phosphorylations in cancers

The paper predicts destabilizing tyrosine phosphorylations in cancers using machine learning, finding 58% of recurrent phosphorylations are destabilizing and identifying pathway and network patterns associated with these events.

DARKIN: A zero-shot benchmark for phosphosite-dark kinase association using protein language models

DARKIN evaluates phosphosite-dark kinase associations using zero-shot classification with protein language models, showing ESM, ProtT5-XL, and SaProt perform best.

Enhancing Protein Binding Site Residue Prediction with Graph Neural Networks: Impacts of Cutoff Distance and Feature Selection

This study evaluates how cutoff distance and feature selection affect protein binding site residue prediction using graph neural networks across four eukaryotic organisms.

LncMamba: A Deep Learning Model for lncRNA Localization Prediction Based on the Mamba Model

LncMamba predicts lncRNA subcellular localization using a deep learning framework combining FPN and Mamba networks, revealing motif relationships.

Rapid Identification of Metal Resistance Genes Using an Enhanced ResNet Deep Learning Model Trained on a largely Expanded BacMet-Based Database

The paper expands the BacMet database and trains an enhanced ResNet model to rapidly identify metal resistance genes with high accuracy and speed.

Evaluating the Effectiveness of Parameter-Efficient Fine-Tuning in Genomic Classification Tasks

This study evaluates parameter-efficient fine-tuning in genomic classification and finds it underperforms compared to full and partial fine-tuning methods.

Protein functional site annotation using local structure embeddings

This paper introduces PARSE, a method for protein functional site annotation using local structure embeddings, achieving high accuracy in predicting catalytic residues and novel functions.

Boosting the Predictive Power of Protein Representations with a Corpus of Text Annotations

This study shows that fine-tuning protein models on text annotations improves their predictive power for function prediction tasks, outperforming BLAST in some cases.

ProtSEC: Ultrafast Protein Sequence Embedding in Complex Space Using Fast Fourier Transform

ProtSEC provides fast and accurate protein sequence embeddings using complex numbers and FFT, achieving significant speed and memory improvements over PLMs.

NucEL: Single-Nucleotide ELECTRA-Style Genomic Pre-training for Efficient and Interpretable Representations

NucEL improves genomic representation learning by using ELECTRA-style pre-training with single-nucleotide tokenization, achieving state-of-the-art performance on multiple tasks.

The Impact of Tokenizer Selection in Genomic Language Models

This study evaluates the impact of tokenizer selection on genomic language models, finding that character tokenization outperforms sub-word methods for nucleotide-level tasks, but byte-pair encoding performs better for SARS-CoV-2 variant classification.

PLM-ICE: A Protein Language Model-based Approach for Prediction of Ice nucleating and Antifreeze Proteins

PLM-ICE predicts ice-binding proteins with high precision and sensitivity using ESM-2 embeddings and an MLP, outperforming existing methods in multiple metrics.

ESM-LoRA-Gly: Improved prediction of N- and O-linked glycosylation sites by tuning protein language models with low-rank adaptation (LoRA)

ESM-LoRA-Gly improves prediction of N- and O-linked glycosylation sites by fine-tuning ESM2-3B with LoRA, achieving better accuracy and efficiency.

Multi-Modal Protein Representation Learning with CLASP

CLASP integrates protein sequence, structure, and text-based descriptions using geometric deep learning and contrastive learning to create embeddings that improve zero-shot classification and clustering.

ICEPIC: A Toolkit to Discover Ice Binding Proteins from Sequence

The paper introduces a toolkit using machine learning to predict ice binding protein potential, activity, and expression from sequences, achieving high accuracy and scalability.

Exo-Tox: Identifying Exotoxins from secreted bacterial proteins

Exo-Tox identifies bacterial exotoxins using a specialized predictor trained on bacterial exotoxins and secreted non-toxic proteins, outperforming existing methods with high accuracy.

Protein Function Prediction via Contig-Aware Multi-Level Feature Integration

This paper introduces CAML, a deep learning model that improves protein function prediction by integrating sequence, structure, and contig-level CDS topology features, achieving significant performance gains over existing methods.

FusionProt: Fusing Sequence and Structural Information for Unified Protein Representation Learning

FusionProt integrates sequence and structural information for protein representation learning, improving performance on enzyme classification, gene ontology, and mutation stability prediction tasks.

ET-Pfam: Ensemble transfer learning for protein family prediction

ET-Pfam improves protein family prediction by combining transfer learning and ensemble strategies, achieving lower error rates than individual models and competitors.

HPInet: Interpretable prediction of Host-Pathogen protein-protein Interactions using a transformer-based neural network

HPInet predicts host-pathogen protein-protein interactions using a transformer-based model, achieving high accuracy and identifying key residues from sequence data.

Scaling down protein language modeling with MSA Pairformer

This paper introduces MSA Pairformer, a memory-efficient protein language model that outperforms existing methods in contact prediction and variant effect prediction.

Beyond the Leaderboard: Leveraging Predictive Modeling for Protein-Ligand Insights and Discovery

The paper integrates ligand-protein interaction data into protein representation learning to improve predictive models for drug discovery and biological mechanisms.

Allo-PED: Leveraging protein language models and structure features for allosteric site prediction

AlloPED predicts allosteric sites using protein language models and machine learning, achieving high accuracy on benchmark datasets.

RAG-ESM: Improving pretrained protein language models via sequence retrieval

RAG-ESM improves pretrained protein language models by incorporating homology information through retrieval-augmentation, achieving better performance in sequence prediction tasks.

C3PI: Component Puzzle Protein-Protein Interaction Prediction

C3PI is a deep learning framework for predicting protein-protein interactions that outperforms existing methods, especially on leakage-free datasets.

Pseudo-perplexity in One Fell Swoop for Protein Fitness Estimation

The paper introduces a method to estimate protein fitness using pseudo-perplexity from a single forward pass through a language model, achieving state-of-the-art results on a benchmark.