sequence-to-function modeling

A Willow Sex Chromosome Reveals Convergent Evolution of Complex Palindromic Repeats

The paper identifies large palindromic structures in the sex chromosome of Salix purpurea, showing convergent evolution with mammalian sex chromosomes.

AStarix: Fast and Optimal Sequence-to-Graph Alignment

AStarix provides a fast and optimal algorithm for aligning sequences to genome graphs, outperforming existing methods by 1-2 orders of magnitude.

Functions of olfactory receptors are decoded from their sequence

The paper uses machine learning to predict olfactory receptor-odorant pairs and basal activity from sequence data, identifying key residues that determine selectivity and activity.

Improving protein alignment algorithms using amino-acid hydrophobicities - Applications of TMATCH, A new algorithm

The paper introduces TMATCH, a new protein alignment algorithm that uses amino-acid hydrophobicity to improve identification of remotely related proteins, outperforming FASTA and PSI-BLAST in some cases.

Hiding in plain sight: structure and sequence analysis reveals the importance of the antibody DE loop for antibody-antigen binding.

The paper analyzes the structure and sequence of the antibody DE loop, revealing its role in antigen binding and its variability in HIV-1 antibodies.

Evolutionary context-integrated deep sequence modeling for protein engineering

This paper introduces ECNet, a deep learning model that uses evolutionary context to predict protein variant fitness, outperforming existing methods on multiple datasets.

Seq2DFunc: 2-dimensional convolutional neural network on graph representation of synthetic sequences from massive-throughput assay

The paper introduces Seq2DFunc, a 2D CNN for predicting nucleic acid sequence functions from synthetic data, showing improved performance over 1D CNNs with smaller datasets.

Charge and hydrophobicity are key features in sequence-trained machine learning models for predicting the biophysical properties of clinical-stage antibodies

The paper develops machine learning models to predict antibody biophysical properties using sequence data, finding that charge and hydrophobicity are key features.

TMATCH: A New Algorithm for Protein Alignments using amino-acid hydrophobicities

TMATCH is a dynamic programming algorithm that uses amino-acid hydrophobicity to improve protein sequence alignment by reinforcing favorable diagonal transitions and penalizing unfavorable ones.

Sequence-to-function deep learning frameworks for synthetic biology

The paper introduces STORM, a deep learning model to optimize toehold switches by analyzing sequence-to-function relationships in a dataset of 91,534 toehold switches.

Global atlas tree of natural proteins based on sorted composition vectors

This paper introduces a non-parametric global sequence comparison method using sorted composition vectors to efficiently classify and compare large protein sequence datasets.

Pooled CRISPR Inverse PCR sequencing (PCIP-seq): simultaneous sequencing of retroviral insertion points and the integrated provirus with long reads

PCIP-seq uses long-read sequencing to simultaneously identify retroviral insertion sites and their integrated proviruses, revealing SNPs, structural variants, and viral evolution.

Deep exploration networks for rapid engineering of functional DNA sequences

This paper introduces Deep Exploration Networks (DENs) to generate high-fitness DNA sequences, demonstrating improved performance in engineering RNA isoforms compared to gradient ascent methods.

Binary integer programming for Multiple Sequence Alignment

The paper presents a mathematical model for Multiple Sequence Alignment (MSA) using Genetic Algorithm to find optimal solutions.

Unified rational protein engineering with sequence-only deep representation learning

This paper introduces UniRep, a deep learning-based representation of protein sequences that enables accurate prediction of protein stability and function while reducing engineering costs.

Thrombospondin module 1 domain (TSP1) of the matricellular protein CCN3 shows an atypical disulfide pattern and incomplete CWR layers

The paper determines the crystal structure of the CCN3 TSP1 domain, revealing an atypical disulfide pattern and incomplete CWR layers, and revises TSP1 domain definitions.

B-SIDER: Computational Algorithm for the Design of Complementary β-sheet Sequences

B-SIDER is a computational method for designing complementary β-strands that outperforms existing methods in distinguishing amyloidogenic β-strands.

Multi-State Design of Flexible Proteins Predicts Sequences Optimal for Conformational Change

This paper demonstrates that multi-state protein design using the RECON algorithm predicts sequences optimal for conformational change, showing increased sequence conservation and resemblance to evolutionary sequence space.

A deep learning approach to pattern recognition for short DNA sequences

A convolutional network trained on 16S reads from 13,838 species reaches within 2.0% of training-set Bayes-optimal species accuracy and usually beats k-mer, alignment and Kraken2 on held-out-species genera.

Deep learning regression model for antimicrobial peptide design

A CNN trained on 4050 E. coli peptide MIC records predicts activity, outperformed other AMP classifiers with random negatives, and guided two designed peptides active against three pathogens.

Critical slowing as a biomarker for seizure susceptibility

This study identifies critical slowing as a biomarker for seizure susceptibility using long-term iEEG recordings from epilepsy patients.

Evaluating Protein Transfer Learning with TAPE

Across five protein tasks with evolution-aware splits, self-supervised pretraining helped nearly every model on every task, yet alignment-based features still beat learned embeddings on secondary structure and contact prediction.

Looking for Therapeutic Antibodies in Next Generation Sequencing Repositories

The paper compares therapeutic antibodies to NGS-derived antibody sequences to identify sequence matches and convergence between natural and artificial antibody development.

Trevolver: simulating non-reversible DNA sequence evolution in trinucleotide context on a bifurcating tree

Trevolver simulates non-reversible DNA sequence evolution with trinucleotide context dependence on bifurcating trees, enabling testing of neutral evolutionary hypotheses across multiple biological contexts.

VSEPRnet: Physical structure encoding of sequence-based biomolecules for functionality prediction: Case study with peptides

The paper introduces VSEPRnet, a method to encode biomolecule structures for functionality prediction, demonstrating improved accuracy in predicting peptide-MHC binding affinities.

Protein Multiple Alignments: Sequence-based vs Structure-based Programs

The paper evaluates sequence- and structure-based multiple alignment programs using reference databases, finding structure-based methods perform better at low sequence identity and for structured residues.

Revealing evolutionary constraints on proteins through sequence analysis

The paper identifies evolutionary sectors in protein sequences by analyzing covariance matrix eigenvalues, linking them to physical properties and mutational effects.

Bioinformatics analysis for L polymerase homology of arenaviruses

The paper uses bioinformatics analysis to identify similarities between arenavirus L polymerase and other hemorrhagic fever viruses in GenBank.

SESCA: Predicting Circular Dichroism Spectra from Protein Molecular Structures

SESCA predicts circular dichroism spectra from protein structures, outperforming existing methods with 50% smaller average deviation.

Sources of suboptimality in a minimalistic explore-exploit task

The paper examines human behavior in a simplified explore-exploit task and finds deviations from optimal strategies, revealing a suboptimal switching rule and sequence-level variability.