sequence-to-function modeling

Machine learning reveals sequence-function relationships in family 7 glycoside hydrolases

This study uses machine learning to identify sequence features that predict functional subtypes and CBM presence in GH7 glycoside hydrolases with high accuracy.

ENNAVIA is an innovative new method which employs neural networks for antiviral and anti-coronavirus activity prediction for therapeutic peptides

The paper introduces ENNAVIA, a deep learning model for predicting antiviral and anti-coronavirus peptide activity, achieving high accuracy on experimental data.

Improving sequence-based modeling of protein families using secondary structure quality assessment

The paper introduces two scoring functions to assess the quality of secondary structures in generated protein sequences and shows improved detection of non-functional sequences.

Puffaligner: An Efficient and Accurate Aligner Based on the Pufferfish Index

PuffAligner is a fast and accurate sequence aligner that balances speed, memory usage, and alignment quality, offering performance comparable to Bowtie2 and STAR.

Learning to Predict in Networks with Heterogeneous and Dynamic Synapses

The paper shows that recurrent neural networks with short-term synaptic plasticity outperform standard models in predictive tasks by leveraging learned synaptic time constants and activity history encoding.

A platform for deep sequence-activity mapping and engineering antimicrobial peptides

The paper develops a platform to map antimicrobial peptide activity and identifies potent mutants through high-throughput sequencing.

Refining pairwise sequence alignments of membrane proteins by the incorporation of anchors

This paper introduces position anchoring in AlignMe to improve pairwise sequence alignment of membrane proteins by incorporating functional or structural information, enhancing accuracy while maintaining reproducibility.

SubFeat: Feature Subspacing Ensemble Classifier for Function Prediction of DNA, RNA and Protein Sequences

SubFeat predicts functionalities of DNA, RNA, and protein sequences using an ensemble classifier, outperforming existing methods on four datasets.

PepVAE3: variational autoencoder framework for antimicrobial peptide generation and activity prediction

PepVAE3 generates antimicrobial peptides using a variational autoencoder and predicts their activity based on sequence and MIC data, enabling controlled design of novel AMPs with experimental validation.

"Multi-Agent" Screening Improves the Efficiency of Directed Enzyme Evolution

This paper introduces a multi-agent directed evolution approach that identifies hidden mutations by using multiple substrates, improving enzyme engineering efficiency.

A comparison of approaches to scaffolding multiple regions along the 16S rRNA gene for improved resolution

This paper compares methods for scaffolding multiple regions of the 16S rRNA gene to improve taxonomic resolution, finding that Sidle provides the least bias and most precise species-level resolution.

Expanding functional protein sequence space using generative adversarial networks

This paper introduces ProteinGAN, a GAN that generates functional protein sequences, showing 24% of generated sequences are soluble and active in vitro.

The genotype-phenotype landscape of an allosteric protein

Dose-response curves for 62 472 LacI coding sequences train a recurrent network for EC50 and G∞, enable EC50 targeting within 1.25-fold, and reveal band-stop sensors.

Syncmers are more sensitive than minimizers for selecting conserved k-mers in biological sequences

The paper introduces syncmers as an alternative to minimizers for selecting conserved k-mers, showing they are more sensitive and robust to mutations.

Generating novel protein sequences using Gibbs sampling of masked language models

The paper generates novel protein sequences using Gibbs sampling of BERT-style masked language models and evaluates their quality against natural sequences.

SAINT: automatic taxonomy embedding and categorization by Siamese triplet network

SAINT is an alignment-free sequence comparison method using a Siamese triplet network for efficient and accurate taxonomy embedding and categorization.

Real-Time, Direct Classification of Nanopore Signals with SquiggleNet

SquiggleNet classifies nanopore signals in real time with high accuracy, outperforming traditional methods in speed and memory usage.

Sequence and structure comparison of ATP synthase F0 subunits 6 and 8 in notothenioid fish.

The paper compares the sequence and structure of ATP synthase subunits 6 and 8 in notothenioid fish with seven other vertebrates to identify physiologically relevant differences.

MotiMul: A significant discriminative sequence motif discovery algorithm with multiple testing correction

MotiMul discovers significant sequence motifs with multiple testing correction, improving statistical power and reducing type-1 error rates.

Aligning biological sequences by exploiting residue conservation and coevolution

The paper introduces DCAlign, a method for sequence alignment that incorporates coevolution and second-order interactions, improving upon profile models for proteins and RNA.

Global analysis of multi-mutants to improveprotein function

This paper introduces GMMA, a method to identify beneficial amino acid substitutions in proteins using multi-mutant data, and applies it to GFP variants to find stabilizing and functional improvements.

Minimally-overlapping words for sequence similarity search

The paper introduces a sparse-seeding method for sequence similarity search using minimally-overlapping words to improve sensitivity compared to existing methods.

Remote homology search with hidden Potts models

This paper introduces a hidden Potts model for sequence alignment and homology search, showing promising results in RNA structure homology benchmarks.

Inferring protein sequence-function relationships with large-scale positive-unlabeled learning

This paper develops a positive-unlabeled learning framework to infer sequence-function relationships from DMS data and applies it to design stabilized enzymes.

Optimizing Consensus Generation Algorithms for Highly Variable Amino Acid Sequence Clusters

The paper introduces a novel consensus generation algorithm for highly variable amino acid sequences, showing it preserves epitopes and maintains antigenicity compared to existing methods.

abPOA: an SIMD-based C library for fast partialorder alignment using adaptive band

abPOA is a fast partial order alignment library using SIMD acceleration, achieving 15x speed improvement over SPOA with similar accuracy.

Protein Structural Alignments From Sequence

The paper introduces DeepBLAST, a method using protein language model embeddings to generate explicit sequence alignments, improving speed and accuracy for structurally similar proteins.

Phylogenetic origin and sequence features of MreB from the wall-less swimming bacteria Spiroplasma

The paper analyzes MreB sequences from Spiroplasma to determine their phylogenetic origin and sequence features, identifying key differences from Bacillus MreBs that may relate to their unique swimming mechanism.

Sapling: Accelerating Suffix Array Queries with Learned Data Models

Sapling improves sequence alignment speed by using a learned data model to enhance suffix array queries, outperforming existing methods on various genomes.

Fast and flexible design of novel proteins using graph neural networks

A deep graph neural network trained on Sudoku puzzles designs novel protein sequences that adopt predetermined structures.