sequence-to-function modeling

Scalable Differentiable for mRNA Design

This paper introduces a scalable differentiable folding algorithm for mRNA design, improving performance and enabling optimization via deep learning.

Evolutionary constraints guide AlphaFold2 prediction of alternative conformations and inform rational mutation design

This paper improves AlphaFold2's ability to predict alternative protein conformations by integrating clustering strategies and evolutionary signals, enabling rational mutation design.

Epileptiform activity and seizure risk follow long-term non-linear attractor dynamics

This paper uses HAVOK analysis to model nonlinear seizure dynamics and improves multi-day seizure risk forecasting by identifying short-period forcings.

Protein Sequence Modelling with Bayesian Flow Networks

This paper introduces ProtBFN, a Bayesian Flow Network for protein sequence generation, which outperforms existing models and is fine-tuned for antibody generation.

Interpretable deep learning reveals the sequence rules of Hippo signaling

The paper uses interpretable deep learning to identify sequence rules governing Hippo signaling in mouse trophoblast stem cells, revealing how transcription factors influence enhancer activity.

Solution structure of the Z0 domain from transcription repressor BCL11A sheds light on the sequence properties of protein-binding zinc-fingers

The paper determines the solution structure of the Z0 domain of BCL11A and identifies sequence patterns distinguishing DNA-binding from non-binding zinc-fingers.

A Simple Generative Model for the Prediction of T-Cell Receptor - Peptide Binding in T-Cell Therapy for Cancer

This paper introduces a deep learning model called GRIP that predicts T-cell receptor epitope peptides, achieving high accuracy and generating novel sequences for cancer immunotherapy.

Combining transformer and 3DCNN models to achieve co-design of structures and sequences of antibodies in a diffusional manner

The paper introduces AlphaPanda, an algorithm combining transformer, 3DCNN, and diffusion models for antibody sequence and structure design, achieving high-quality results with reduced training data.

Investigating Data Size, Sequence Diversity, and Model Complexity in MPRA-based Sequence-to-Function Prediction

The paper creates a curated MPRA dataset collection and builds sequence-to-function models to analyze how data size, diversity, and model complexity affect predictive performance.

Inference and visualization of complex genotype-phenotype maps with gpmap-tools

gpmap-tools uses Gaussian processes to infer and visualize genotype-phenotype maps with higher-order epistasis; across 262,144 Shine-Dalgarno variants it reveals register-specific peaks linked by three-nucleotide-offset ridges.

Seq2Topt: a sequence-based deep learning predictor of enzyme optimal temperature

Seq2Topt predicts enzyme optimal temperature using protein sequences with deep learning, achieving high accuracy and identifying key residues for temperature adaptation.

Ali-U-Net: A Convolutional Transformer Neural Net for Multiple Sequence Alignment of DNA Sequences. A proof of concept

Ali-U-Net is a convolutional transformer neural network that aligns DNA sequences more accurately and faster than existing tools in specific scenarios.

EVOFLOW-RNA: GENERATING AND REPRESENTING NON-CODING RNA WITH A LANGUAGE MODEL

EvoFlow-RNA generates and represents non-coding RNA using a masked discrete diffusion model, outperforming existing models on RNA function tasks.

Genome modeling and design across all domains of life with Evo 2

A 40B-parameter DNA model trained on 9.3 trillion tokens scores coding and noncoding variants zero-shot, exposes interpretable sequence features, and writes chromosome-scale sequences.

TIPs-VF: An augmented vector-based representation for variable-length DNA fragments with sequence, length, and positional awareness

TIPs-VF provides a variable-length DNA sequence encoding framework that improves representation of genetic sequences for machine learning by incorporating sequence, length, and positional awareness.

Characterization of Soybean Vein Necrosis Virus (SVNV) Proteins: Sequence Analysis of Field Strains and Comparison of Localization Patterns in Differing Cell Types

This study characterizes SVNV proteins through sequence analysis of field strains and compares their localization in plant and insect cells.

Amino acid properties, substitution rates, and the nearly neutral theory

The paper tests if amino acid properties affect substitution rates, finding that larger physicochemical differences between amino acids correlate with lower substitution rates across multiple datasets.

DDS-E-Sim: A Transformer-based Probabilistic Generative Framework for Simulating Error-Prone DNA Sequences for DNA Data Storage

DDS-E-Sim simulates error-prone DNA sequences for DNA data storage, capturing both random and biased errors with high fidelity across different sequencing technologies.

Sidechain conditioning and modeling for full-atom protein sequence design with FAMPNN

This paper introduces FAMPNN, a method for protein sequence design that explicitly models both amino acid identity and sidechain conformation, achieving state-of-the-art results in sidechain packing and practical design applications.

Physics-inspired accuracy estimator for model-docked ligand complexes

The paper introduces DENOISer, a deep neural network that improves scoring accuracy in model docking by incorporating physical knowledge, outperforming existing tools on benchmark sets.

Geometric Protein Optimization

This paper introduces Geometric Protein Optimization (GPO), an AI framework that improves protein-ligand binding affinity by optimizing global geometry through diverse substitutions, achieving significant in silico improvements.

GRATCR: epitope-specific T cell receptor sequence generation with data-efficient pre-trained models

GRATCR generates epitope-specific TCR sequences with fewer training data than state-of-the-art models, showing higher specificity and biological functionality.

Beyond RNA Structure Alone: Complex-Aware Feature Fusion for Tertiary Structure-based RNA Design

The paper introduces CARD, a method that integrates complex-level information to improve tertiary structure-based RNA design by incorporating protein features and affinity evaluation.

Sequence similarity estimation by random subsequence sketching

This paper introduces SubseqSketch, an alignment-free method for sequence similarity estimation that uses random subsequences to efficiently approximate edit distances, showing effectiveness in benchmark tasks.

AnchoRNA: Full virus genome alignments through conserved anchor regions

AnchoRNA improves full-genome alignment of viral sequences by identifying conserved anchor regions, enhancing alignment accuracy and efficiency.

A ML Framework for Genetic Sequence Identification using 2D Electrical Conductance Probability Distributions from Mixed Data Sets

The paper proposes a machine learning framework combining CNN and XGBoost to identify genetic sequences from 2D electrical conductance probability distributions, showing a 10% performance improvement.

JRSeek: Artificial Intelligence Meets Jelly Roll Fold Classification in Viruses

JRSeek uses logistic regression on LLM embeddings to predict jelly roll folds in viral sequences with over 95% accuracy, showing the feasibility of sequence-based prediction.

Hybrid Generative Model: Bridging Machine Learning and Biophysics to Expand RNA Functional Diversity

The paper introduces a hybrid generative model for RNA design that combines coevolutionary likelihood with secondary structure priors to expand functional sequence diversity.

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.