Bioinformatics Advances

review_benchmark/figure at main · zlsys3/review_benchmark Benchmark testing of cloning evolutionary inference tools on simulated data - zlsys3/review_benchmark

📊 Code and data used for figure generation are openly available here:

Benchmarking of seven methods highlights strengths and weaknesses, and a novel spatiotemporal framework is proposed linking phylogenetic branch lengths with spatial transcriptomic gradients.

This review assesses over 20 tools for #tumor #phylogenetic inference across cross-sectional, regional bulk, single-cell, and lineage tracing designs.

🧪 Just out in Bioinformatics Advances: “Computational strategies in tumor phylogenetics: Evaluating multi-modal integration and methodological trade-offs across study designs”  

Explore the full study: https://doi.org/10.1093/bioadv/vbaf242

netrias (Netrias) harmonization, standardization, curation, ontology alignment, data generation, large language models

💻 Resources including training, validation, and test datasets, along with representative GPT-2 models, are openly available via the Netrias Hugging Face organization (https://huggingface.co/netrias).

Using data augmentation to mimic real-world term variations, the models achieved 96% in-dictionary accuracy and substantially reduced manual standardization effort compared to heuristics and zero-shot GPT-4o.

The study presents fine-tuned GPT-2 models for harmonizing inconsistent metadata across domains such as cancer, alcohol research, and infectious disease.

🗂️ Now published in Bioinformatics Advances: “Metadata harmonization from biological datasets with language models”

Read the full paper here: https://doi.org/10.1093/bioadv/vbaf241

🧭 The authors emphasize emerging directions including DNA language models, integration of comparative genomics and transcriptomic data, and improved benchmarking frameworks to advance accurate and robust gene prediction across diverse eukaryotic species.

This review synthesizes eukaryotic gene prediction methods, proposing a taxonomy by gene-model reliance (gene-model-based, gene-model-free, hybrid). It covers classical and #deeplearning approaches, extrinsic evidence sources, and identifies key strengths, limitations, and challenges.

📚 Explore the latest from Bioinformatics Advances: “An overview of computational methods for gene prediction in eukaryotes: Strengths, limitations, and future directions”   

Full article available: https://doi.org/10.1093/bioadv/vbaf222

GitHub - CMUSchwartzLab/TUSV-INT Contribute to CMUSchwartzLab/TUSV-INT development by creating an account on GitHub.

💻 Code available at

TUSV-int integrates bulk DNA-seq and scRNA-seq within an integer linear programming framework to jointly model SNVs, CNAs, and SVs. Benchmarks on simulated and real #breastcancer data show improved clonal deconvolution and #phylogeny inference over existing methods.

🧬 Explore the latest from Bioinformatics Advances: "Deconvolution and phylogeny inference of diverse variant types integrating bulk DNA-seq with single-cell RNA-seq"

Full article available: https://www.doi.org/10.1093/bioadv/vbaf234

Cytoscape App Store - SocialViruses

💻 Open source Cytoscape app available here:

SocialViruses is a Cytoscape application for rational phage cocktail design. It incorporates quantitative phage–bacteria and phage–phage interaction networks, supports up to 12 phages, minimizes antagonism and redundancy, and provides detailed performance metrics across diverse datasets.

🦠 Now published in Bioinformatics Advances: "SocialViruses: Integrating quantitative phage–bacteria and phage–phage interaction networks for rational cocktail design" 

Read the full paper here: https://doi.org/10.1093/bioadv/vbaf239

GitHub - yuyiwen-yiyuwen/Disc_Hub: a software package for evaluating machine learning performance a software package for evaluating machine learning performance - yuyiwen-yiyuwen/Disc_Hub

📦 Open source code available here:

Disc-Hub benchmarks 3 training strategies and 4 classifiers on DIA-MS datasets, showing that K-fold training with multilayer perceptrons best balances identification depth and FDR control. The package enables rapid, reproducible evaluation of #machinelearning configurations for DIA identification.

🧪 Just out in Bioinformatics Advances: "Disc-Hub: a python package for benchmarking machine learning strategies in DIA-MS identification" 

Explore the full study: https://www.doi.org/10.1093/bioadv/vbaf232

GitHub - Mordziarz/TAILcaller: an R package for analyzing polyA tails after dorado basecalling. an R package for analyzing polyA tails after dorado basecalling. - Mordziarz/TAILcaller

🛠️ Software available open source:

TAILcaller is an R package designed to analyze poly(A) tail length differences directly from dorado-generated BAM files. It supports both direct RNA and cDNA nanopore sequencing data, enabling global, gene-level, and transcript-level analyses with flexible statistical testing and visualization.

🧬 Just out in Bioinformatics Advances: "TAILcaller: An R package for analyzing differences in poly(A) tail length for Oxford Nanopore RNA sequencing” 

Full article available: https://doi.org/10.1093/bioadv/vbaf235

GitHub - MaayanLab/sc2DAT Contribute to MaayanLab/sc2DAT development by creating an account on GitHub.

🛠️ Try sc2DAT here: https://sc2dat.maayanlab.cloud/  
Source code available:

sc2DAT is a web-based workflow that integrates single-cell and bulk RNA-seq data to automatically identify cell subpopulations, rank cell-surface targets, and predict therapeutic compounds. It leverages resources like LINCS L1000 and TargetRanger for drug and target prioritization.

🧬 Now published in Bioinformatics Advances: “sc2DAT: Workflow for targeting tumor subpopulations of single cells”

Full article available: https://doi.org/10.1093/bioadv/vbaf237

GitHub - epifluidlab/FinaleToolkit: FinaleToolkit is a package and standalone program to extract fragmentation features of cell-free DNA from paired-end sequencing data. FinaleToolkit is a package and standalone program to extract fragmentation features of cell-free DNA from paired-end sequencing data. - epifluidlab/FinaleToolkit

🧰 Open-source toolkit with CLI, Python API, and documentation available:

Benchmarks on >1B fragments confirmed accuracy, scalability, and memory efficiency.

FinaleToolkit is a Python package for efficient extraction of cfDNA fragmentation features. It replicates >10 published fragmentation metrics, supports parallel processing, and achieves up to 50-fold faster performance than original implementations.

🧪 Now published in Bioinformatics Advances: "FinaleToolkit: Accelerating cell-free DNA fragmentation analysis with a high-speed computational toolkit"

Explore the full study: https://doi.org/10.1093/bioadv/vbaf236