sequence-to-function modeling

RecAlign: A* recombination-aware sequence to graph mapping

RecAlign improves recombination-aware sequence-to-graph mapping by using an A* algorithm with haplotype information and a heuristic function, achieving faster and more efficient performance.

GALILEO Generatively Expands Chemical Space and Achieves One-Shot Identification of a Library of Novel, Specific, Next Generation Broad-Spectrum Antiviral Compounds at High Hit Rates

This study uses GALILEO and ChemPrint to generate novel antiviral compounds targeting the conserved Thumb-1 site, achieving 100% hit rate in vitro against HCV and Coronavirus 229E.

Simulating 500 million years of evolution with a language model

ESM3, a 98-billion-parameter multimodal protein language model tokenized across sequence, structure, and function, designs a bright green fluorescent protein at 58% sequence identity to known fluorescent proteins.

Simultaneous polyclonal antibody sequencing and epitope mapping by cryo electron microscopy and mass spectrometry - a perspective.

The paper evaluates using cryoEM and mass spectrometry with a deep-learning tool to sequence antibodies and map epitopes from convalescent serum, showing 80-90% accuracy in variable domain sequencing.

Effective prediction of IL-17 inducing peptides using hybrid approach: iIL17pred

This paper introduces a hybrid method for classifying IL-17 inducing peptides, achieving higher accuracy and MCC than existing methods.

Improving the generalization of protein expression models with mechanistic sequence information

This study improves sequence-to-expression models by integrating mechanistic sequence features with one-hot encodings, enhancing generalization across Escherichia coli and Saccharomyces cerevisiae datasets.

Prediction of protein biophysical traits from limited data: a case study on nanobody thermostability through NanoMelt

NanoMelt, a ridge-stacked ensemble of regressors on four diverse sequence embeddings, predicts nanobody melting temperature from 640 curated sequences, reaching Spearman 0.832 under nested cross-validation.

AllerTrans: A Deep Learning Method for Predicting the Allergenicity of Protein Sequences

This paper introduces AllerTrans, a deep learning model that predicts protein allergenicity from sequences, achieving high sensitivity and specificity.

DDGemb: predicting protein stability change upon single- and multi-point variations with embeddings and deep learning

DDGemb predicts protein stability changes from single- and multi-point variations using embeddings and deep learning, achieving state-of-the-art performance.

UTRGAN: Learning to Generate 5' UTR Sequences for Optimized Translation Efficiency and Gene Expression

UTRGAN generates 5' UTR sequences optimized for higher translation efficiency and gene expression, outperforming existing UTRs in multiple metrics.

ChromBPNet: bias factorized, base-resolution deep learning models of chromatin accessibility reveal cis-regulatory sequence syntax, transcription factor footprints and regulatory variants

Learning and removing Tn5/DNase-I sequence bias lets a small base-resolution CNN (ChromBPNet) recover TF motifs and footprints from accessibility data and rival Enformer at predicting chromatin QTL effects.

BioStructNet: Structure-Based Network with Transfer Learning for Predicting Biocatalyst Functions

BioStructNet predicts biocatalyst functions using structure-based deep learning and transfer learning, improving accuracy for small datasets.

Generation of antigen-specific paired chain antibody sequences using large language models

This paper introduces MAGE, a model that generates paired human antibody sequences against specific targets, validated against SARS-CoV-2, H5N1, and RSV-A.

Kun-peng: an ultra-memory-efficient, fast, and accurate pan-domain taxonomic classifier for all

Kun-peng is a metagenomic classifier that outperforms existing tools in speed, memory efficiency, and accuracy for large-scale pan-domain classification.

Leveraging ancestral sequence reconstruction for protein representation learning

This paper introduces LASE, a method using ancestral sequence reconstruction to train protein language models with higher accuracy and faster computation than large models.

OPUS-GO: An interpretable protein/RNA sequence annotation framework based on biological language model

OPUS-GO provides sequence-level and residue-level annotations for proteins and RNA sequences using a biological language model and modified MIL strategy, improving accuracy and interpretability.

End-to-end simulation of nanopore sequencing signals with feed-forward transformers

The paper introduces seq2squiggle, a transformer-based model for simulating nanopore sequencing signals, which outperforms existing tools on real data.

RNA language models predict mutations that improve RNA function

The paper develops an RNA language model to predict mutations that enhance RNA thermostability using a new database of RNA sequences linked to environmental temperatures.

A Large-Scale Foundation Model for RNA Function and Structure Prediction

This paper introduces AIDO.RNA, a large-scale foundation model for RNA function and structure prediction, achieving state-of-the-art performance on multiple tasks.

HyperMPNN - A general strategy to design thermostable proteins learned from hyperthermophiles

This paper introduces HyperMPNN, a deep learning model trained on hyperthermophile proteins to design highly thermostable proteins, achieving stability up to 95°C.

Limitations and Enhancements in Genomic Language Models: Dynamic Selection Approach

This paper proposes a dynamic model selection approach for genomic language models, showing improved performance through multi-model fusion and identifying limitations related to motif reliance.

ESM-Scan - a tool to guide amino acid substitutions

ESM-Scan uses ESM zero-shot prediction to identify amino acid substitutions in proteins, showing effectiveness in predicting stability and function changes through benchmarking and experimental validation.

Multi-purpose controllable protein generation via prompted language models

This paper introduces PROPEND, a protein design method using prompted language models to directly control multiple properties, achieving higher functional recovery than traditional pipelines.

Sequence modeling and design from molecular to genome scale with Evo

7B-parameter convolution-attention DNA model with 131 kb context, trained on prokaryotic genomes, performs zero-shot fitness and essentiality prediction and generates Cas loci, transposons, and 650 kb sequences.

A Rapid, Cost-Effective Method for High-Yield DNA Purification from PCR Products and Agarose Gels.

The paper develops a cost-effective in-house DNA purification protocol for PCR products and agarose gels, showing improved clarity and purity compared to commercial kits.

TCR-TRANSLATE: Conditional Generation of Real Antigen Specific T-cell Receptor Sequences

TCR-TRANSLATE generates antigen-specific TCR sequences using seq2seq models, achieving validated CDR3β sequences for unseen antigens but with limitations in validation scope and focus.

Sequencing Strategy to Ensure Accurate Plasmid Assembly

The paper evaluates sequencing strategies for accurate plasmid assembly, finding that hybrid approaches provide the most reliable results while highlighting challenges in GC-rich and repetitive regions.

Metapredict enables accurate disorder prediction across the Tree of Life

Metapredict V3 predicts intrinsically disordered regions across 20,000 proteomes to assess disorder prevalence across life kingdoms.

TSTA: Thread and SIMD-Based Trapezoidal Pairwise/Multiple Sequence Alignment Method

The paper introduces TSTA, a parallel sequence alignment method that improves efficiency using thread and SIMD parallelism for pairwise and multiple sequence alignments.

A long-context language model for deciphering and generating bacteriophage genomes

This paper introduces a long-context language model for bacteriophage genomes, capable of predicting gene functions and generating novel sequences.