seqeron-mirna

name	seqeron-mirna
description	microRNA (miRNA) analysis with Seqeron (MCP tools OR the C# API). Use for canonical miRNA target-site prediction (8mer / 7mer-m8 / 7mer-A1 / 6mer, Bartel/TargetScan hierarchy) with AU/positional context and local structure accessibility, seed extraction / seed-family grouping / similar-miRNA search / single-nucleotide seed variants, miRNA→target antiparallel duplex alignment, pre-miRNA hairpin detection (mature/star arms + dot-bracket + Turner energy), and RNA base-pairing / G-U wobble tests. Triggers: "find miRNA target sites", "predict miRNA targets in this 3'UTR/mRNA", "what's the seed of this miRNA", "miRNA seed family", "group these miRNAs by seed", "find similar miRNAs", "generate seed variants", "align this miRNA to its target", "find pre-miRNA hairpins", "is this a G-U wobble pair in a miRNA duplex", "can these two miRNA/target bases pair", "reverse-complement a seed to get its target motif", "miRNA target accessibility / context score". A context-free base-pairing / G-U question with no miRNA framing → seqeron-rna-structure. Server: annotation (all MiRnaAnalyzer.*). THREE guarded units — see the Envelope section.
allowed-tools	Read, Bash, Grep, Glob

seqeron-mirna — miRNA targets, seeds/families, pre-miRNA hairpins, wobble pairing

Routing + orchestration skill for the microRNA family on the annotation server (14 tools, one backing class: MiRnaAnalyzer.*). It picks the right tool for a target-site / seed-family / pre-miRNA / pairing question and gives a dual-mode recipe (MCP tool name and the equivalent Seqeron.Genomics C# Method ID). This is the densest guarded family in the repo — three guarded units — so the ⚠ Envelope section below is the point of the skill.

• Rigor is delegated. Tool-only computation, provenance, cross-check, units/0-based coords, and the alpha / not-for-clinical caveat are owned by bio-rigor — applies by default; do not restate.
• Don't know a tool name? Use seqeron-discovery (python3 scripts/skills/find-tool.py <kw> --server annotation) — never guess.
• Point, don't duplicate. Full I/O schemas: docs/mcp/tools/annotation/*.md; algorithm invariants: docs/algorithms/MiRNA/*.md.

Scope boundary — this family vs neighbours

• This skill OWNS the miRNA workflow: canonical target-site prediction + context + accessibility, seed extraction / families / variants / similarity, miRNA→target duplex alignment, pre-miRNA hairpin (biogenesis precursor) detection, and miRNA base-pairing / wobble. bio-annotation name-drops the miRNA row shallowly — route here.
• Pure RNA secondary-structure folding (MFE, general stem-loops, pseudoknots, dot-bracket) → seqeron-rna-structure. Disambiguation: a structural hairpin/stem-loop question ("fold this RNA", "find stem-loops") is rna-structure; a miRNA-biogenesis precursor hairpin ("find pre-miRNA hairpins", mature/star arms) is find_pre_mirna_hairpins here. MiRnaAnalyzer.CanPair / IsWobblePair are the miRNA copies (distinct Method IDs from RnaSecondaryStructure.CanPair).
• mRNA 3'UTR / gene structure / ORFs / splicing → bio-annotation.

Decision guide — which tool for which question

Question	Tool ([MCP] / Method ID)
Build a miRNA record (T→U, seed 2–8 extracted)	create_mirna / MiRnaAnalyzer.CreateMiRna
Just the seed (positions 2–8)	mirna_seed_sequence / MiRnaAnalyzer.GetSeedSequence
Scan an mRNA for canonical target sites (8mer/7mer/6mer)	find_mirna_target_sites / MiRnaAnalyzer.FindTargetSites ⚠
AU / positional context around a site	analyze_target_context / MiRnaAnalyzer.AnalyzeTargetContext ⚠
Local structure accessibility of a site	site_accessibility / MiRnaAnalyzer.CalculateSiteAccessibility
Antiparallel miRNA↔target duplex (matches/wobbles/ΔG)	align_mirna_to_target / MiRnaAnalyzer.AlignMiRnaToTarget
Group miRNAs by identical seed (families)	group_by_seed_family / MiRnaAnalyzer.GroupBySeedFamily
Pairwise seed comparison (same family?)	compare_seed_regions / MiRnaAnalyzer.CompareSeedRegions
Find miRNAs within a seed-Hamming budget	find_similar_mirnas / MiRnaAnalyzer.FindSimilarMiRnas
Enumerate single-substitution seed variants	generate_seed_variants / MiRnaAnalyzer.GenerateSeedVariants
Detect pre-miRNA hairpins (mature/star arms)	find_pre_mirna_hairpins / MiRnaAnalyzer.FindPreMiRnaHairpins ⚠
Seed's target pattern (reverse complement)	rna_reverse_complement / MiRnaAnalyzer.GetReverseComplement
Can two bases pair (A-U, G-C, G-U)?	can_pair / MiRnaAnalyzer.CanPair
Is a pair a G-U wobble?	is_wobble_pair / MiRnaAnalyzer.IsWobblePair

⚠ = touches a guarded unit — read the Envelope section before relying on the output.

Canonical dual-mode pipelines

(a) Predict miRNA target sites in an mRNA (main entry)

1. [MCP] create_mirna(name, sequence) → record (T→U, seed 2–8). (MiRnaAnalyzer.CreateMiRna)
2. [MCP] find_mirna_target_sites(mRnaSequence, miRna, minScore?=0.5) → sites[{start,end,type,score,freeEnergy,…}] (0-based). (.FindTargetSites)
3. [MCP] cross-check a hit's seed pattern independently: rna_reverse_complement(seed) should equal the targetSequence seed match. (.GetReverseComplement)

• ⚠ MIRNA-TARGET-001: the returned score is base site-type + pairing only, not the full context++ score. See Envelope.

(b) Context + accessibility scoring of a found site

1. [MCP] analyze_target_context(mRnaSequence, targetStart, targetEnd, contextWindow?=30) → auContent, nearStart/nearEnd, contextScore∈[0,1]. (.AnalyzeTargetContext)
2. [MCP] site_accessibility(mRnaSequence, siteStart, siteEnd) → accessibility∈[0,1] (±50 nt window). (.CalculateSiteAccessibility)

• ⚠ MIRNA-TARGET-001: contextScore is a partial context proxy (AU + position), not TargetScan context++. See Envelope.

(c) Seed families / similarity / variants

1. [MCP] group_by_seed_family(miRnas) → families[{seedFamily,members}]. (.GroupBySeedFamily)
2. [MCP] compare_seed_regions(miRna1, miRna2) → matches, mismatches, isSameFamily. (.CompareSeedRegions)
3. [MCP] find_similar_mirnas(query, database, maxMismatches?=1) → seed-Hamming neighbours. (.FindSimilarMiRnas)
4. [MCP] generate_seed_variants(seedSequence) → original + all 3L single substitutions. (.GenerateSeedVariants)

• No guarded unit here — seed handling is exact string logic.

(d) miRNA↔target duplex alignment

1. [MCP] align_mirna_to_target(miRnaSequence, targetSequence) → alignmentString (|/:/ ), matches, guWobbles, freeEnergy (Turner 2004, kcal/mol). (.AlignMiRnaToTarget)
2. [MCP] spot-check a column: can_pair(b1,b2) / is_wobble_pair(b1,b2). (.CanPair / .IsWobblePair)

• No guarded unit — antiparallel (miRNA i vs target[len-1-i]), ungapped.

(e) Pre-miRNA hairpin detection

1. [MCP] find_pre_mirna_hairpins(sequence, minHairpinLength?=55, maxHairpinLength?=120, matureLength?=22) → hairpins[{start,end,sequence,matureSequence,starSequence,structure,freeEnergy}]. (.FindPreMiRnaHairpins)

• ⚠ MIRNA-CLEAVAGE-001 (star-arm span) and MIRNA-PRECURSOR-001 (no read-support score). See Envelope.

⚠ Envelope — THREE guarded units (STOP rules)

Authoritative source: docs/Validation/LIMITATIONS.md. Do not raise LimitationPolicy to force output (bio-rigor rule 2) — STOP and report + name the alternative.

• MIRNA-TARGET-001 — MinimumMode Permissive (guarded; throws in Strict & Moderate). Scope limit: a full context++ score is not produced out-of-the-box — find_mirna_target_sites gives site type + pairing only; analyze_target_context gives a partial AU+position proxy. TA_3UTR, PCT, SPS, Len_ORF, ORF8m are all caller-supplied (no default transcriptome / TargetScan sigmoid tables are bundled). STOP rule: if asked for a TargetScan-grade context++ score, either supply a 3'UTR set (then ComputeTa3Utr derives TA_3UTR) + alignment/tree/sigmoid params for PCT + SPS/ORF features, or report the limitation and name TargetScan — do not present the partial score as context++.
• MIRNA-CLEAVAGE-001 — MinimumMode Permissive (guarded; throws in Strict & Moderate). Scope limit: the 5p Drosha/Dicer cut reproduces miRBase exactly, but the miRNA*(3p)/star-arm boundary returned by find_pre_mirna_hairpins/cleavage is an approximate linear 2-nt-3′-overhang offset, not the miRBase-annotated span (real boundaries encode dominant sequencing-read cut sites). STOP rule: if the exact star-arm coordinates matter, supply the miRBase mature-3p (MIMAT) coordinates or small-RNA-seq read pileups, or report the limitation — do not treat the offset span as the true annotation.
• MIRNA-PRECURSOR-001 — documented-only, no runtime guard (nothing throws; the shortfall is undetectable per call). Scope limit: the read-stacking / read-support score of miRDeep2 is not implemented (the closed-form log-odds is only in a gated 2008 supplement + GPL source). find_pre_mirna_hairpins returns a structural candidate (fold + energy), not a read-supported confidence. STOP rule: if asked for read-support / miRDeep2-style confidence, report that it is out of scope and name miRDeep2 with the caller's small-RNA-seq reads — do not imply the structural candidate is read-validated.

End-to-end grounded example

Task. Given let-7a and a candidate 3'UTR, find its target sites, score context + accessibility of the top hit, and independently corroborate the seed pattern — flagging every guarded layer.

1. create_mirna("let-7a","UGAGGUAGUAGGUUGUAUAGUU") → seed GAGGUAG. (MiRnaAnalyzer.CreateMiRna)
2. find_mirna_target_sites(mRnaSequence, miRna, minScore=0.5) → sites (0-based; type, score). (.FindTargetSites) — ⚠ score is site-type+pairing (MIRNA-TARGET-001).
3. rna_reverse_complement("GAGGUAG") → CUACCUC; confirm it appears in the top site's targetSequence (independent cross-check). (.GetReverseComplement)
4. analyze_target_context(mRnaSequence, top.start, top.end, 30) → auContent, contextScore. (.AnalyzeTargetContext) — ⚠ partial context (MIRNA-TARGET-001).
5. site_accessibility(mRnaSequence, top.start, top.end) → accessibility. (.CalculateSiteAccessibility)

Provenance
1) create_mirna(name=let-7a,sequence=UGAGGUAGUAGGUUGUAUAGUU) → seed=GAGGUAG
2) find_mirna_target_sites(mRnaSequence,miRna,minScore=0.5) → sites[{start,end,type,score,freeEnergy}] (0-based)
3) rna_reverse_complement(GAGGUAG)=CUACCUC ≡ top-site seed match (cross-check)
4) analyze_target_context(mRnaSequence,top.start,top.end,contextWindow=30) → auContent,contextScore
5) site_accessibility(mRnaSequence,top.start,top.end) → accessibility
Envelope: MIRNA-TARGET-001 — score/context are partial, NOT TargetScan context++ (supply 3'UTR set/PCT params or cite TargetScan).
Caveat: alpha software; not for clinical use — independently validate before relying on any call.

Reference

• This family's tool map (all 14 — curated index; NOT in domain-map.json, so there is NO generated slice): reference/tool-map.md
• Fuller recipes + parameter guidance: reference/pipelines.md
• Algorithm background (invariants/formulas — link, don't copy): Seed_Sequence_Analysis.md · Target_Site_Prediction.md · MiRNA_Target_Pairing.md · Pre_miRNA_Detection.md
• Operating envelope / guarded units: LIMITATIONS.md (MIRNA-TARGET-001, MIRNA-CLEAVAGE-001, MIRNA-PRECURSOR-001)
• Cross-cutting: bio-rigor (rigor guardrail) · seqeron-discovery (tool lookup) · bio-annotation (parent — it name-drops the miRNA row; route here) · seqeron-rna-structure (pure RNA folding / stem-loops vs miRNA-biogenesis hairpins — disambiguate).