sequence-to-function modeling

In silico design of DNA sequences for in vivo nucleosome positioning

The paper designs synthetic DNA sequences for in vivo nucleosome positioning in yeast using a kinetic Monte Carlo framework combined with deep learning predictions.

Likelihood-based fine-tuning of protein language models for few-shot fitness prediction and design

This paper compares likelihood-based and embedding-based methods for fitness prediction and design using protein language models, finding likelihood-based fine-tuning consistently outperforms embedding-based approaches.

Cliffy: robust 16S rRNA classification based on a compressed LCA index

Cliffy improves 16S rRNA classification by reducing index size and increasing accuracy compared to Kraken2 and Bracken.

Designing DNA With Tunable Regulatory Activity Using Discrete Diffusion

This paper introduces D3, a generative model for designing regulatory DNA with tunable activity, showing improved performance over baselines in sequence generation and activity prediction.

A generative foundation model for antibody sequence understanding

FAbCon is a generative antibody-specific language model that improves antigen binding prediction and antibody design with increasing model size.

Designing meaningful continuous representations of T cell receptor sequences with deep generative models

This paper introduces TCR-VALID, a variational autoencoder for generating and analyzing TCR sequences, showing improved clustering and representation quality.

An integrated technology for quantitative wide mutational scanning of human antibody Fab libraries

The paper introduces MAGMA-seq, a technology for quantitative analysis of antibody sequence-function relationships, demonstrating its application in mapping antibody development pathways and identifying binding determinants.

Predicting transcriptional activation domain function using Graph Neural Networks

The paper predicts transcriptional activation domain function using graph neural networks, achieving high accuracy and outperforming existing methods.

Ancestral reconstruction of polyethylene terephthalate degrading cutinases reveals a rugged and unexplored sequence-fitness landscape

This study uses ancestral sequence reconstruction to explore PET-degrading cutinase sequence space, revealing diverse evolutionary pathways and the effectiveness of LASE in mapping fitness landscapes.

Centrifuger: lossless compression of microbial genomes for efficient and accurate metagenomic sequence classification

Centrifuger compresses microbial genomes losslessly for faster and more accurate metagenomic classification by reducing memory use and improving taxonomic resolution.

For antibody sequence generative modeling, mixture models may be all you need

This paper introduces a simple generative model, SAM, for antibody sequence modeling that outperforms complex models and provides fast, interpretable tools for antibody sequence analysis.

RiboDiffusion: Tertiary Structure-based RNA Inverse Folding with Generative Diffusion Models

RiboDiffusion generates RNA sequences from 3D structures using a diffusion model, outperforming baselines in sequence recovery and showing consistent performance across RNA types and lengths.

Arrayed in vivo barcoding for multiplexed sequence verification of plasmid DNA and demultiplexing of pooled libraries

This paper introduces an in vivo plasmid barcoding method for high-throughput sequence verification and demultiplexing of pooled libraries, demonstrating high accuracy and cost-effectiveness.

Learning the language of proteins and predicting the impact of mutations

This paper introduces a protein language model that predicts mutation impacts and lineage classification in viral genomes, offering faster and more accurate functional predictions than traditional methods.

DeepMicroClass sorts metagenomes into prokaryotes, eukaryotes and viruses, with marine applications

DeepMicroClass classifies metagenomic contigs into five sequence classes with high accuracy, outperforming existing methods in marine metagenomic analysis.

gRNAde: Geometric Deep Learning for 3D RNA inverse design

gRNAde designs RNA sequences based on 3D backbone structures using geometric deep learning, achieving higher native sequence recovery than Rosetta on benchmark tasks.

Protein language model powers accurate and fast sequence search for remote homology

PLMSearch uses a protein language model to perform fast and accurate homologous protein sequence search, improving sensitivity over traditional methods while matching structure search performance.

Effect of Tokenization on Transformers for Biological Sequences

This study evaluates how different tokenization methods affect transformer models' performance on various biological sequence tasks, showing improved accuracy and efficiency with optimized tokenizers.

Design of Antigen-Specific Antibody CDRH3 Sequences Using AI and Germline-Based Templates

This paper designs antigen-specific antibody CDRH3 sequences using AI and germline-based templates, validated against SARS-CoV-2.

Perturbation-aware predictive modeling of RNA splicing using bidirectional transformers

The paper trains bidirectional transformer models to predict RNA splicing using perturbative RNA data, achieving state-of-the-art performance in splice junction prediction.

All-atom protein sequence design based on geometric deep learning

This paper introduces GeoSeqBuilder, a geometric deep learning framework for protein sequence design that achieves high accuracy in recovering native residues and predicting side-chain conformations.

RNAinformer: Generative RNA Design With Tertiary Interactions

RNAinformer designs RNAs from secondary structures by inverting folding algorithms, including non-canonical and tertiary interactions, achieving state-of-the-art performance.

PPFlow: Target-aware Peptide Design with Torsional Flow Matching

PPFlow designs target-aware peptides using torsional flow matching, achieving state-of-the-art performance in peptide drug generation and optimization.

Riboformer: A Deep Learning Framework for Predicting Context-Dependent Translation Dynamics

Riboformer predicts context-dependent translation dynamics using deep learning, correcting artifacts and identifying sequence motifs affecting ribosome stalling.

Accurate prediction of site- and amino-acid substitution rates with a mutation-selection model

This study introduces a mutation-selection model to predict site- and amino-acid substitution rates from amino acid sequences, showing it outperforms standard phylogenetic methods and correlates with empirical Bayes estimates.

Harnessing Transformers to Generate Protein Sequences Prone to Liquid Liquid Phase Separation

This paper uses GPT-based models to generate protein sequences prone to liquid-liquid phase separation and validates their ability to form phase-separated states.

Sequence-function mapping of proline-rich antimicrobial peptides

The paper maps sequence-function relationships in proline-rich antimicrobial peptides using SAMP-Dep, identifying mutational patterns and epistasis that influence activity and specificity.

PandoGen: Generating complete instances of future SARS-CoV-2 sequences using Deep Learning

PandoGen generates future SARS-CoV-2 protein sequences using deep learning, forecasting novel variants with higher accuracy than larger models.

ProteinMPNN Recovers Complex Sequence Properties of Transmembrane β-Barrels

This study benchmarks ProteinMPNN for designing transmembrane β-barrel proteins, showing it outperforms Franklin2023 with refined inputs and generates sequences suitable for experimental validation.

Target-aware Molecule Generation for Drug Design Using a Chemical Language Model

This paper introduces TamGen, a chemical language model for target-aware molecule generation, and shows it produces high-quality compounds with inhibitory activity against Tuberculosis ClpP protease.