Changelog

v2.4.0 — September 26, 2026

MCP for AI agents, faster spatial processing, friendlier tables

MCP for AI agents — expressRNA now exposes a read-only MCP server, so an AI assistant like Claude can look up your samples, libraries, analyses and results directly, and explain exactly how a result was computed instead of just reading back numbers. Result tables can be filtered by significance and gene name, so a large analysis doesn't get cut off arbitrarily. Get your access token from your profile page.
Spatial Patterns refinements — domain maps are now grouped by condition with a shared color scale and new fill display modes, per-domain effect sizes are shown directly in the results table, detection thresholds are editable, and a data-import bug with headered position files is fixed.
Import spatial samples from GEO/ENA — paste an accession and expressRNA fetches the FASTQ and position files directly, matching the existing bulk-sample import workflow.
Friendlier data tables — sample, library, CLIP and analyses tables are now resizable and reorderable with show/hide columns; gene tables on the APA, DGE and iCLIP pages are paginated with instant search instead of one long scroll; every table and edit page now shows its record ID.
Faster, more reliable background processing — polyA site counting is roughly 10x faster, the analysis worker automatically restarts if it crashes, and a mapping/counting bug affecting some spatial and MDS results is fixed.
Homepage refresh — the Capabilities section is now icon tiles split into core and newer features, including the new MCP integration.

v2.3.0 — September 20, 2026

Genome Browser, RNAmotifs2 cluster motif analysis

Every mapped sample now has a Genome Browser tab (built on igv.js) so you can inspect the actual aligned reads, search by gene or jump to a locus, and share the current view as a link — multi-gigabyte CRAM files open instantly via HTTP range requests, no download needed.

RNAmotifs2 cluster motif analysis is now integrated: it searches for clusters of short RNA motifs enriched around regulated splice junctions or polyA sites, with Fisher's exact test and FDR correction. It auto-runs on every finished APA analysis, and a new standalone analysis type lets you upload your own event table.

August 2026

Expression Time Course & multi-section Spatial Patterns

Expression Time Course plots one or more genes across a library's samples in your own order — developmental stages, a time series, treatment steps — with a "draw a profile" search (find genes matching a hand-drawn pattern) and live GO enrichment.

Spatial Patterns extends Spatial APA to compare multiple Visium sections at once: sections are jointly clustered into shared spatial domains, then tested for differential patterns between two groups in both gene expression and polyA site usage.

July 2026

Alternative splicing analysis & spatial APA

Bulk RNA-seq analyses now automatically trigger an alternative splicing pipeline (powered by splicekit) once differential expression is complete, with junction/exon/gene-level results browsable in the analysis view.

Spatial APA is now supported for Visium-style datasets — polyA site usage quantified per barcode directly from BAM files, producing spatially resolved maps comparable across tissue sections.

June 2026

Platform revival & full recode

expressRNA returns with a complete ground-up rewrite: a lightweight Python/SQLite backend and a modern single-page jQuery interface. Highlights: redesigned UI, Google OAuth login with groups and visibility controls, a new Samples upload flow, a revamped Libraries module with polyAsite reference integration, Alternative Polyadenylation analyses with RNA-Map plots and CLIP integration, and a background worker for automated processing.

polyAI: predict polyA cleavage sites in human sequence with deep learning

polyAI is a free online tool that predicts where a messenger RNA is cleaved and polyadenylated, from DNA sequence alone. Paste a sequence or FASTA file (up to 1,000,000 nt) or type a human gene name, and a dilated convolutional neural network trained on 1.6 billion read 3' ends from 428 public Oxford Nanopore direct-RNA sequencing runs (GRCh38) returns a per-base cleavage density track, the called sites ranked by strength, their poly(A) signal hexamers, and, for a gene, the observed polyAtlas sites for comparison. Predicted summits fall within 2 nt of the PolyASite 2.0 reference position for 35% of clusters and within 10 nt for 62%. No login is needed.

polyAI is part of expressRNA, an open platform for post-transcriptional RNA regulation, and a companion of polyAtlas, the polyA site atlases it was trained on.