Machine Learning in Bioinformatics

Genolator is a multimodal large language model that combines genomic and structural data with natural language to interpret protein function, outperforming existing models in accuracy.

TooTranslator uses language embedding alignment to predict substrates of transport proteins, achieving 80% accuracy with top-100 predictions for unseen substrates.

This study evaluates an embedding-based protein sequence alignment method using clustering and double dynamic programming, showing improved performance in detecting remote homology compared to traditional and state-of-the-art approaches.

The paper introduces Profluent-E1, retrieval-augmented protein language models that improve performance by conditioning on homologous sequences, achieving state-of-the-art results in fitness prediction and contact-map prediction.

SignalGen predicts optimal signal peptides for protein expression by integrating mature protein sequences, organism, and localization using a Latent Residual Transformer.

This paper introduces ESMSCOP, a contrast-aware pre-training framework that integrates evolutionary and structural information for improved protein function prediction.

The paper introduces a zero-shot protein segmentation method using ProtT5 embeddings to identify and categorize functional protein segments without sequence conservation reliance.

The paper improves prediction of functional protein interface sequences by integrating structural and sequence features using a graph attention model, achieving better generalization across diverse variants.

MotifAE discovers functional sequence patterns from protein language models through an unsupervised framework, improving motif identification and enabling stability prediction.

The paper identifies 15 candidate cytoplasmic retention peptides in human proteins, with 10 having unrelated sequences, and shows that retention sequences are located at specific sites across protein isoforms.

MkAtt-SDN2GO improves protein function prediction by integrating multi-kernel convolution and attention mechanisms, achieving better performance than SDN2GO on Gene Ontology terms.

PLiCat uses large language models to classify protein-lipid interactions across eight lipid categories and predict lipid-binding sites and mutation impacts.

ProAffinity++ predicts protein-protein binding affinities using sequence and structure representations, outperforming existing methods in accuracy and handling complex cases like antigen-antibody interactions.

This study benchmarks fourteen protein-language models (pLMs) to determine which best captures biological properties, finding that mid-scale models often outperform larger ones for general tasks.

ProteinSight predicts carotenoid-binding sites using a 3D U-Net model, showing high sensitivity and specificity on a curated test set and identifying potential sites on Human Serum Albumin.

DomDiff improves domain boundary identification and classification using a diffusion model with ESM2 embeddings, outperforming existing methods and excelling in rare families.

The paper introduces SaprotHub, a platform that simplifies training and sharing of protein language models, enabling non-experts to build and collaborate on customized models.

StoPred predicts protein complex stoichiometry using protein language models and graph attention networks, achieving higher accuracy for both homomeric and heteromeric complexes than existing methods.

AstraPTM2 predicts 39 PTM types on full-length proteins using a transformer model with structural and sequence features, achieving high AUROC and macro-F1 scores.

The paper develops a pipeline to annotate protein functions using latent features from protein language models, identifying missing annotations and structurally matching proteins.

ProTrek introduces a tri-modal contrastive learning model for protein sequence, structure, and function, achieving significant improvements in speed and accuracy over existing tools.

Qimai improves DNA-protein interaction prediction for unseen proteins by integrating deep learning with LLM-based reasoning, achieving higher AUC-PR, AUC-ROC, and MCC than standalone models.

This study uses protein language models and Bayesian optimization to improve protein engineering efficiency by predicting function from sequences and optimizing variant search.

This paper introduces InterPLM, a framework using sparse autoencoders to extract interpretable features from protein language models, revealing both known and novel biological concepts.

DeepGene is a foundation model for genomics that addresses genetic language diversity, model efficiency, and length extrapolation, achieving top performance on 9 out of 28 genome understanding tasks.

This study fine-tunes pre-trained protein language models on viral protein sequences to improve their performance on downstream tasks and representation quality.

PlasmoFP uses deep learning to predict protein function in Plasmodium species, reducing unannotated proteins from 15-59% to 3-28% and increasing fully annotated proteins from 7-42% to 36-68%.

This paper identifies High Attention (HA) sites in protein sequences using the ESM model to improve protein family classification and function prediction.

This paper introduces a curriculum learning approach for training antibody language models, showing that a 650M-parameter model outperforms existing mixed models in downstream tasks.

The paper introduces a framework that integrates DNA sequences with omics data using a pretrained model, enabling reconstruction and unified representation of genomic and functional information.