Machine Learning in Bioinformatics

LOCAS: Multi-label mRNA Localization with Supervised Contrastive Learning

LOCAS improves mRNA subcellular localization by using supervised contrastive learning and a multi-label classification head, achieving state-of-the-art performance.

Protein Language Models: Is Scaling Necessary?

This paper introduces AMPLIFY, a protein language model that achieves better performance with less training cost by improving data quality and architecture, challenging the need for scaling.

Prediction of anti-freezing proteins from their evolutionary profile

This paper develops a machine learning model to predict antifreeze proteins using evolutionary information, achieving higher accuracy than existing methods on an independent dataset.

Distinguishing word identity and sequence context in DNA language models

The paper evaluates DNA language models by assessing their ability to distinguish sequence context from token identity, finding that overlapping k-mer tokenization limits context learning.

Prediction of Protein Half-lives from Amino Acid Sequences by Protein Language Models

The paper develops a protein half-life prediction model, PLTNUM, using a protein language model trained on NIH3T3 mouse data, achieving 71% accuracy and identifying key sequence features affecting half-life.

Interpretable improving prediction performance of general protein language model by domain-adaptive pretraining on DNA-binding protein

This paper introduces ESM-DBP, a domain-adapted protein language model for DNA-binding proteins, showing improved prediction performance and interpretability.

AutoPeptideML: A study on how to build more trustworthy peptide bioactivity predictors

This paper develops AutoPeptideML, an AutoML tool for peptide bioactivity prediction, showing that automated methods can improve model trustworthiness and efficiency.

Protein domain embeddings for fast and accurate similarity search

The paper introduces DCTdomain, a method that segments proteins into domains and uses discrete cosine transformation to create domain-level embeddings for improved protein similarity search.

PEFT-SP: Parameter-Efficient Fine-Tuning on Large Protein Language Models Improves Signal Peptide Prediction

PEFT-SP improves signal peptide prediction using parameter-efficient fine-tuning on large protein language models, achieving higher MCC2 gains with limited training data.

Are Genomic Language Models All You Need? Exploring Genomic Language Models on Protein Downstream Tasks

This study evaluates genomic language models on protein tasks, showing they are competitive with proteomic models and benefit from joint genomic-proteomic training and optimized tokenization.

Fine-tuning protein language models boosts predictions across diverse tasks

This study shows that fine-tuning protein language models improves prediction performance across tasks and can be done efficiently, especially for small datasets.

Predicting the bacterial host range of plasmid genomes using the language model-based one-class SVM algorithm

This paper introduces HRPredict, a method using one-class SVM and language model-based word vectors to predict plasmid host ranges across bacterial taxa, achieving high accuracy in benchmark tests.

MuLAN: Mutation-driven Light Attention Networks for investigating protein-protein interactions from sequences

MuLAN predicts protein interactions and binding affinity changes from sequence data using a deep learning approach without requiring structural information.

Multi-Modal Large Language Model Enables Protein Function Prediction

ProteinChat is a multi-modal large language model that predicts protein functions from amino acid sequences, outperforming general-purpose LLMs and task-specific models.

Prediction of Oligomeric Status of Quaternary Protein Structure by using Sequential Minimal Optimization for Support Vector Machine

The paper predicts the oligomeric status of proteins using a support vector machine with sequential minimal optimization based on amino acid composition.

The OMG dataset: An Open MetaGenomic corpus for mixed-modality genomic language modeling

The Open MetaGenomic (OMG) corpus combines 3.1T base pairs and 3.3B protein coding sequences in mixed amino-acid and nucleotide format to pretrain the genomic language model gLM2.

TooT-PLM-P2S: Incorporating secondary structure information into protein language models

This paper introduces TooT-PLM-P2S, a protein language model incorporating secondary structure information, and evaluates its performance on eight downstream tasks.

FEDKEA: Enzyme function prediction with a large pretrained protein language model and distance-weighted k-nearest neighbor

FEDKEA improves enzyme function prediction by combining a pretrained protein language model with distance-weighted k-nearest neighbor, outperforming existing tools.

Fine-tuning of conditional Transformers for the generation of functionally characterized enzymes

This paper introduces Finenzyme, a Protein Language Model that improves EC category prediction through fine-tuning, generating functionally similar enzymes with preserved structure and function.

UniPTM: Multiple PTM site prediction on full-length protein sequence

UniPTM predicts multiple PTM sites on full-length protein sequences without using sliding windows, outperforming existing models and providing interpretable results.

Annotation Vocabulary (Might Be) All You Need

The paper introduces Annotation Vocabulary to enhance protein embeddings using structured ontologies, demonstrating its effectiveness through various downstream tasks and generative models.

A hybrid approach for predicting transcription factors

This study develops a hybrid method combining alignment-based and alignment-free approaches to predict transcription factors with high accuracy, achieving an AUC of 0.99 on an independent dataset.

DeepEnzyme: a robust deep learning model for improved enzyme turnover number prediction by utilizing features of protein 3D structures

DeepEnzyme improves kcat prediction using Transformer and GCN with sequence and 3D structure features, showing robustness for low similarity enzymes and mutation impact analysis.

PLM_Sol: predicting protein solubility by benchmarking multiple protein language models with the updated Escherichia coli protein solubility dataset

This study develops PLM_Sol, a protein solubility predictor using PLMs and an updated E. coli dataset, achieving improved accuracy, F1 score, and MCC on independent and experimental test sets.

AAontology: An ontology of amino acid scales for interpretable machine learning

AAontology classifies 586 amino acid scales into 8 categories and 67 subcategories to improve interpretability in protein prediction tasks.

Beware of Data Leakage from Protein LLM Pretraining

This study evaluates data leakage effects in protein language model pretraining on thermostability prediction by comparing dataset split strategies.

ToxinPred 3.0: An improved method for predicting the toxicity of peptides

This paper introduces ToxinPred 3.0, an improved method for predicting peptide toxicity using alignment-free machine/deep learning and ensemble approaches, achieving high accuracy and sensitivity.

HPClas: A data-driven approach for identifying halophilic proteins based on catBoost

HPClas is a machine learning classifier using catBoost to identify halophilic proteins with high accuracy, achieving AUROC scores of 0.877 and 0.845 on training and test sets respectively.

PMTPred: Machine Learning Based Prediction of Protein Methyltransferases using the Composition of k-spaced Amino Acid Pairs

The paper introduces PMTPred, a machine learning tool for predicting protein methyltransferases using sequence-based features, achieving high accuracy on benchmark datasets.

ProtParts, an automated web server for clustering and partitioning protein datasets

ProtParts clusters protein sequences using BLAST E-values and graph algorithms to create independent partitions, improving model performance by reducing overfitting.