🧪 The memod-s workflow is available at
13.03.2026 09:02
👍 0
🔁 1
💬 0
📌 0
memod-s is a Snakemake-based workflow that streamlines analysis of prokaryotic DNA methylation from Nanopore sequencing data. The pipeline integrates basecalling, quality control, genome assembly, annotation, & methylation profiling to generate genome-wide methylation statistics & visualisations.
🧬 New in Bioinformatics Advances: "Memod-s: A standardised workflow to explore and analyse prokaryotic methylation patterns for Nanopore sequencing data"
Explore the study: https://doi.org/10.1093/bioadv/vbag072
Authors include: @alessiome.bsky.social, @jaimemurtaza.bsky.social
🤖 miRInter-Trans is available at https://github.com/AnacletoLAB/miRInter-Trans.
miRInter-Trans is a transformer-based framework for predicting microRNA interactions directly from RNA sequences. It combines embeddings from the RNA-FM foundation model with a feed-forward neural network to detect interactions across miRNA–lncRNA, miRNA–miRNA, and miRNA–snoRNA datasets.
🧬 New in Bioinformatics Advances: "Mirinter-Trans: A transformer-based framework for microRNA interaction prediction"
Read the article: https://doi.org/10.1093/bioadv/vbag073
🧰 OmniBio is available at https://sdehnhardt.shinyapps.io/OmniBio_beta2/. Source code is available at
OmniBio is a web application for calculating microbial growth kinetics from microplate reader data. Built in R using gcplyr and Shiny, it processes raw plate reader outputs and experimental metadata to compute parameters such as lag time, maximum growth rate, and ODmax.
🧫 New in Bioinformatics Advances: "OmniBio: An easy-to-use web app for kinetic growth parameter calculation from microplate reader data"
Read the paper: https://doi.org/10.1093/bioadv/vbag074
📊 Across ontologies, embedding-based neighborhoods achieved substantially higher functional agreement than domain-content similarity alone, and identified significantly more non-sharing architecture pairs with GO similarity greater than 0.8 than expected by chance.
Embeddings more accurately group architectures by shared GO annotations and reveal functionally similar pairs that share no domains.
This study applies NLP-inspired vector embeddings to model human multidomain protein architectures. Using TF-IDF, PMI, and Word2Vec representations of domain sequences, the authors compare embedding-based proximity with Jaccard domain similarity.
🧩 Now published in Bioinformatics Advances: "Vector semantics of multidomain protein architectures"
Read the full paper here: https://doi.org/10.1093/bioadv/vbag037
The approach achieves near state-of-the-art tracking accuracy and substantially improves lineage assignment over nearest-cell heuristics.
A graph neural network models cell geometry and neighborhood relationships for frame-to-frame assignment, while a fully connected network infers mother-daughter pairs from dynamic growth and orientation features.
This study introduces LYN-track and LYN-trace, deep learning methods for tracking and lineage tracing in densely packed budding yeast colonies.
🧫 Explore the latest from Bioinformatics Advances: "Love-thy-neighbor: Neural networks for tracking and lineage tracing in budding yeast"
Full article available: https://doi.org/10.1093/bioadv/vbag067
Authors include: @sophiemartinlab.bsky.social
💻 The prediction server is available at https://web.iitm.ac.in/bioinfo2/PANDriver/index.html. Source code at https://github.com/medhapandey63/PANDriver.git.
The models achieve strong balanced performance across diverse cancers and outperform several existing variant effect predictors.
This study develops cancer-type specific deep neural network models to distinguish driver and passenger missense mutations across 30 tumor types. It integrates sequence-derived features, AlphaFold structural properties, and amino acid network metrics to capture mutation context.
🧬 Just out in Bioinformatics Advances: "Classification of driver and passenger mutations in different cancer types using deep neural networks"
Read the full paper here: https://doi.org/10.1093/bioadv/vbag068
The approach improves peptide-MHC, peptide-TCR, and full-complex prediction while enabling cross-species generalization.
TCRLens is a structure-aware deep learning framework for TCR-pMHC-I recognition. It models 5 interface zones using residue-level graphs and an equivariant graph neural network, augmented with VAE-GAN–generated weak-binding decoys.
🧪 Now published in Bioinformatics Advances: "TCRLens: Structure-aware equivariant graph learning for TCR-pMHC-I recognition and immunogenic epitope discovery"
Read the full paper here: https://doi.org/10.1093/bioadv/vbag066
💻 THe Biom is available at https://thebiom.compbio.ulaval.ca/. Source code at
It integrates stage-specific biomarkers, Reactome pathway annotations, and expression data, enabling cross-stage and pan-cancer comparison. The atlas currently includes 57 signatures spanning 6 tumor types.
THe Biom is an interactive web platform for exploring transcriptomic gene signatures derived from hybrid ensemble feature selection across TCGA cancers.
🧬 Explore the latest from Bioinformatics Advances: "THe Biom: A platform for visualization and exploration of cancer transcriptomic biomarkers identified by robust feature selection"
Full article available: https://doi.org/10.1093/bioadv/vbag065
Authors include: @drarno.bsky.social
