miRPRISM

Batch query

Paste gene symbols (or miRNA IDs in miRNA mode), one per line or comma-separated. Up to 200. Each returns its top hits.

Top per query

About miRPRISM

miRPRISM — microRNA Prioritization by Regression-informed Integrated Sequence-expression Modeling.

miRNA regulation is quantitative and context-dependent: the same candidate site is repressed to different degrees depending on whether the 3′UTR region is actually expressed in a given cellular state. Like a prism resolving mixed light into a spectrum, miRPRISM resolves context-specific repression that static sequence tools cannot see.

The model integrates the 3′UTR sequence, its per-base RNA-seq coverage, and the miRNA sequence through base-pairing attention, and was trained on a confidence-scored, context-resolved atlas of repression strength (RS) derived by bootstrap elastic-net regression of target mRNA on miRNA and transcription-factor expression. It reproduces expression-derived repression rankings on held-out interactions and generalises to an independent paired cohort.

Predictions reflect an expression-derived repression measure, not an experimental causal ground truth. Genome build hg38.

How to use

Reference mode (no upload): pick a gene and get miRNAs ranked by predicted repression strength against a pooled human reference coverage. Zero setup.
Upload mode: provide your sample's 3′UTR RNA-seq coverage as a bigWig (hg38) to get a context-specific ranking.

Generating a bigWig from your BAM

bamCoverage -b sample.bam -o sample.bw \
    --skipNonCoveredRegions

Genome build must be hg38. Coverage is normalised and processed server-side. BAM files are not accepted (too large); bigWig is compact and indexed.

Context-resolved miRNA repression-strength prediction

3′UTR coverage & binding sites