feed

SpatialAgent is an autonomous LLM-driven agent that spans the full spatial biology research workflow — experimental design, multimodal spatial transcriptomics/scRNA-seq analysis, and hypothesis generation — using 72 specialized tools and 17 guided skill templates. It runs in autonomous or co-pilot mode and matched or outperformed human scientists on benchmark tasks across ~2 million cells from brain, heart, and colitis-model tissue.

SPATIA is a multimodal deep learning framework from the Harvard MIMS lab that jointly models cell imagery and gene expression to predict and generate spatial cell phenotypes. It uses cross-attention to fuse morphology with transcriptomics and transformer modules to aggregate niche- and tissue-level context, and can generate cell morphology images conditioned on predicted state transitions via flow matching.

OpenProteo is an open-source Rust library (with Python bindings and a CLI) that reads proprietary mass-spectrometry raw formats from Thermo, Bruker, and Waters instruments and converts them into standard mzML, without requiring vendor SDKs or Windows-only DLLs. It's an umbrella project pairing with sibling crates OpenTimsTDF (Bruker timsTOF) and OpenWRaw (Waters MassLynx).

FLAMES v2.6.0 (Bioconductor 3.23) is a modular, protocol-agnostic R/Bioconductor package for full-length isoform analysis from long-read single-cell and spatial RNA-seq data. It supports droplet-based and combinatorial barcoding single-cell methods as well as spatial transcriptomics workflows, and scales to multi-sample cohort analyses. Key outputs include isoform detection, alternative splicing quantification, and differential splicing analysis integrated with Bioconductor's genomic infrastructure.

fRagmentomics characterizes cell-free DNA fragments overlapping somatic mutations from BAM files, supporting multiple data formats and mutation representation conventions. It is designed for liquid biopsy and cancer genomics applications. New in Bioconductor 3.23.

MetaboAnnotatoR performs feature annotation on LC-MS all-ion fragmentation (AIF) datasets using fragment ion libraries, enabling automated metabolite identification in untargeted metabolomics workflows. New in Bioconductor 3.23.

RFantibody (Nature 2025) is an integrated pipeline for structure-based de novo antibody and nanobody design, combining RFdiffusion (backbone generation), ProteinMPNN (CDR sequence design), and fine-tuned RoseTTAFold2 (structure validation). Cryo-EM confirms atomic accuracy of designed VHHs targeting influenza haemagglutinin and C. difficile toxin B. Free for academic and commercial use.

sc-SPLASH extends the SPLASH statistical framework to barcoded scRNA-seq (10x Chromium) and spatial transcriptomics (10x Visium), enabling reference-free, statistics-first discovery of transcriptomic variation without a reference genome. Its standalone BKC preprocessing submodule is ~50× faster than UMI-tools, and the pipeline has demonstrated discovery of novel secreted proteins in non-model organisms, tumor-associated mutations, and spatially-regulated alternative splicing. Published in Nature Biotechnology (April 2026).

Cardinal is a mass spectrometry imaging toolbox for R providing efficient preprocessing, spatial segmentation, and statistical classification of imaging MS datasets. It serves workflows in proteomics, lipidomics, and spatial metabolomics research. Released in Bioconductor 3.23 (v3.14.0).

FRASER (Find Rare Aberrant Splicing Events in RNA-seq) detects rare aberrant splicing events in transcriptome data using an autoencoder to model read-count expectations and a beta-binomial distribution to score outliers. It supports alternative splicing and intron-retention detection and is optimised for rare disease diagnosis pipelines. Released in Bioconductor 3.23 (v2.8.0).

AlphaDIA is an open-source, feature-free DIA (data-independent acquisition) proteomics search engine that performs machine learning directly on raw mass spectrometry signals, achieving end-to-end transfer learning across instruments without requiring spectral library features. Benchmarked at >120,000 precursors per run, it matches or exceeds other leading search engines. Published in Nature Biotechnology (2025); part of the AlphaX proteomics ecosystem from the Mann Lab.

pertpy is a newly promoted scverse core package dedicated to single-cell perturbation experiment analysis, including CRISPR screens and compound treatments. It provides differential perturbation analysis, signature scoring, and dose-response modeling with purpose-built metadata handling and visualizations. Integrates natively with AnnData and the full scverse stack.

Novae is a graph-based foundation model for spatial transcriptomics data, published in Nature Methods (2025) and integrated into the scverse ecosystem. It is a self-supervised graph attention network enabling zero-shot or fine-tuned spatial domain identification, native batch-effect correction across technologies, and spatially variable gene and pathway analysis. Trained on ~30 million cells across 18 tissues from Xenium, MERSCOPE, and CosMX platforms.

Nicheformer is a transformer-based foundation model published in Nature Methods (2025, vol. 22, pp. 2525–2538), pretrained on SpatialCorpus-110M — a curated corpus of over 57 million dissociated and 53 million spatially resolved cells across 73 tissues from human and mouse. The model enables prediction of spatial context for dissociated scRNA-seq cells, spatial composition prediction, and label transfer via linear probing and fine-tuning. Key finding: models trained only on dissociated data fail to recover spatial microenvironment complexity, underscoring the need for joint single-cell and spatial pretraining.