sequence-to-function modeling

RNA2PS: A sequence-specific coarse-grained RNA model linking structure, thermodynamics and phase separation

RNA2PS predicts RNA structure and phase separation behavior from sequence, showing sequence-dependent duplex stability and condensation differences in trinucleotide repeat RNAs.

GenNA: Conditional generation of nucleotide sequences guided by natural-language annotations

GenNA generates nucleotide sequences using natural-language annotations, showing improved performance in semantic and species context alignment without explicit biological rule supervision.

yakRNA Design: A semantic multimodal RNA composer

The paper introduces yakRNA Design, a language model that generates functional synthetic RNAs without 3D structural data, achieving efficient frame-shifting capabilities in ribosomes.

Refining sequence-to-expression modelling with chromatin accessibility

The paper improves sequence-to-expression models by incorporating chromatin accessibility, showing better performance in predicting gene expression and cell type-specific patterns.

Spatial-neighbour encoding enables fast RNA 3D structure search

The paper introduces RiboSeek, a framework for RNA 3D structure search using geometric alphabets, achieving fast and accurate comparisons.

A trainable language model for modulating translation rates in non-model organisms by generating upstream untranslated region sequence libraries

The paper generates synthetic 5'-UTR sequence libraries for non-model organisms using a language model trained on genomic statistics, showing effective translation modulation without prior knowledge.

Modular Deep Learning for Direct RNA Sequence Design via Self-Contained RNA Units

The paper introduces SCRU-DB, a database of RNA units, and two models, SCRU-Seq and SCRU-Diff, to design RNA sequences with high structural accuracy.

Evolutionary conditioning enables guided generation of functionally diverse enhancers

This paper introduces EnhancAR, a deep learning model that generates functionally diverse enhancers using evolutionary homologs instead of cell type labels.

AlphaFind v2: Similarity Search in AlphaFold DB and TED Domains across Structural Contexts

AlphaFind v2 enables fast, structure-based search across protein structures and domains, using embeddings and US-align for refinement and supporting organism and CATH-based filtering.

PARAS: high-accuracy machine-learning of substrate specificities in nonribosomal peptide synthetases

Random forest predictors PARAS and PARASECT, trained on 3,653 curated adenylation domains, outperform earlier NRPS substrate predictors and correctly assigned L-Trp to an unusual tryptopeptin synthetase domain.

Maximally Divergent Synonymous Gene Design with SIRIUS

SIRIUS generates maximally divergent DNA sequences for a given protein, reducing shared subsequences compared to existing methods.

Separating selection from mutation in antibody language models

This paper introduces a model that separates mutation and selection processes in antibody language models to improve prediction of mutation effects on antibody function.

CREsted: modeling genomic and synthetic cell type-specific enhancers across tissues and species

CREsted is a software package for modeling and designing cell type-specific enhancers using deep learning, validated across tissues and species.

mRNA-GPT: A Generative Model for Full-Length mRNA Design and Optimization

mRNA-GPT is a generative model that optimizes full-length mRNA sequences by jointly optimizing all three regions, achieving Pareto-optimal designs and outperforming existing methods.

DefensePredictor: A Machine Learning Model to Discover Novel Prokaryotic Immune Systems

DefensePredictor combines ESM2 protein language model embeddings with genomic context features in a gradient boosting classifier to identify prokaryotic anti-phage defense genes, validating 45 previously unknown Escherichia coli systems.

DECODING SYNONYMOUS CODON SELECTION WITH A TRANSFORMER MODEL

The paper develops a transformer model to predict codon sequences from amino acids, improving rare codon prediction and linking sequence variation to protein function.

AINN-P1: A Compact Sequence-Only Protein Language Model Achieves Competitive Fitness Prediction on ProteinGym

A Long-Context Generative Foundation Model Deciphers RNA Design Principles

EVA is a long-context generative RNA model trained on 114 million sequences, achieving state-of-the-art performance in RNA design tasks and structural modeling.

Designing mRNA coding sequence via multimodal reverse translation language modeling with Pro2RNA

Pro2RNA generates species-adapted mRNA coding sequences from protein sequences using a multimodal language model, outperforming existing methods in benchmark tests.

SpeciefAI: Multi-species mRNA-level Antibody Framework Generation using Transformers

SpeciefAI generates species-specific mRNA and protein sequences for antibodies and nanobodies using a transformer model trained on six species.

De novo design of therapeutic scFvs and multi-specific engagers from sequence alone

This paper introduces IASO, a sequence-centric framework for designing therapeutic scFvs and multi-specific engagers using only antigen sequence, demonstrating its ability to identify novel binders and overcome drug resistance.

Fast, accurate construction of multiple sequence alignments from protein language embeddings

This paper introduces ARIES, a method using protein language model embeddings to construct multiple sequence alignments, achieving higher accuracy than existing methods, especially in low-identity regimes.

Cyclic peptides space: The methodology of sequence selection to cover the comprehensive physical properties

The paper introduces a methodology to generate cyclic peptide libraries with uniform coverage of physicochemical properties using a protein language model and cyclic permutation averaging.

Generalise or Memorise? Benchmarking Ligand-Conditioned Protein Generation from Sequence-Only Data

This study benchmarks ligand-conditioned protein generation using sequence-only data, revealing trade-offs between diversity and foldability based on dataset characteristics.

Neurotox: Deep learning decodes conserved hallmarks of neurotoxicity across venomous species

Neurotox identifies sequence-based neurotoxicity hallmarks across species using deep learning, showing structural changes correlate with loss of predicted neurotoxicity.

Nanopore-Based Glycan Sequencing via Fragmentation-Reassembly Strategy

This study introduces a nanopore-based method for glycan sequencing that reconstructs full glycan structures from fragments, overcoming limitations of existing approaches.

Assessment of Generative De Novo Peptide Design Methods for G Protein-Coupled Receptors

This paper benchmarks deep learning-based peptide design methods for GPCRs, finding that validation fails to distinguish valid from invalid designs, highlighting scoring challenges.

High Diversity Gene Libraries Facilitate Machine Learning Guided Exploration of Fluorescent Protein Sequence Space

This study shows that expanding training data diversity with synthetic gene libraries improves machine learning models' ability to generate functional fluorescent proteins beyond natural sequence space.

Hybrid Deep Learning with Protein Language Models and Dual-Path Architecture for Predicting IDP Functions

IDPFunNet predicts IDR functions using a hybrid deep learning model, outperforming existing methods in binding prediction and DFL identification across multiple benchmarks.

Contrastive learning for antibody-antigen sequence-to-specificity prediction

This paper introduces CALM, a model that predicts antibody-antigen binding from sequences using contrastive learning, achieving 7% top-1 retrieval on a test set.