| name | seqeron-protein-features |
| description | Predict features of a PROTEIN sequence with Seqeron (MCP tools OR the C# API). Use for intrinsic disorder / IDRs (TOP-IDP) and MoRFs, signal-peptide prediction + cleavage site, transmembrane helices & membrane topology, coiled-coil regions, protein domains, PROSITE / common protein motifs, protein low-complexity (SEG) regions, and Kyte–Doolittle hydrophobicity profiles. Triggers: "predict disorder in this protein", "find the intrinsically disordered regions", "find the signal peptide / where is it cleaved", "transmembrane regions of this protein / membrane topology", "coiled-coil prediction", "what domains / PROSITE motifs are in this protein", "hydrophobicity plot", "is this residue disorder-promoting". Input is an AMINO-ACID sequence. Server: analysis. |
| allowed-tools | Read, Bash, Grep, Glob |
seqeron-protein-features — disorder/IDRs, signal peptide, TM, coiled-coil, domains/motifs, hydrophobicity
Routing + orchestration skill for the protein-feature-prediction family on the Analysis server
(backing classes DisorderPredictor.*, ProteinMotifFinder.*, plus SequenceStatistics for the
hydrophobicity profile). Input is always a protein (amino-acid) sequence. It picks the right tool and
gives a dual-mode recipe (MCP tool call and the equivalent Seqeron.Genomics C# Method ID).
- Rigor is delegated. Tool-only computation, envelope, provenance, cross-check, coordinate reporting,
and the alpha / not-for-clinical caveat are owned by
bio-rigor — it applies
by default; do not restate its rules.
- Don't know a tool name? Use
seqeron-discovery
(python3 scripts/skills/find-tool.py <kw> --server analysis) — never guess.
- Point, don't duplicate. Full I/O schemas live in
docs/mcp/tools/analysis/<tool>.md; algorithm
invariants in docs/algorithms/ProteinPred/ + docs/algorithms/ProteinMotif/. This skill links, it does not copy.
Scope boundary — protein features vs DNA annotation
This skill OWNS the protein-SEQUENCE feature family. The overloaded bio-annotation
name-drops it but its real subject is DNA-shaped annotation. Route by input:
- PROTEIN input → here. Disorder, MoRFs, signal peptide, TM helices, coiled-coils, protein domains,
protein PROSITE / common motifs, protein SEG low-complexity, hydrophobicity.
- DNA/RNA input →
bio-annotation. ORFs/genes/promoters, variants, DNA motifs (exact/degenerate/PWM,
MotifFinder.*), nucleotide low-complexity/DUST (SequenceComplexity.*), repeats, k-mers, splicing,
methylation, RNA structure. Do not confuse ProteinMotifFinder.FindMotifByProsite (protein, here) with
the DNA motif tools there. If you only have DNA, translate to protein first via bio-qc (translate) before entering this skill.
Decision guide — which tool for which question
| Question (protein input) | Tool ([MCP] / Method ID) |
|---|
| Per-residue disorder + IDRs + disorder content (TOP-IDP) | predict_disorder / DisorderPredictor.PredictDisorder ⚠ confidence guarded |
| Molecular-recognition features within disorder | predict_morfs / DisorderPredictor.PredictMoRFs |
| Single amino-acid disorder propensity / is-promoting | disorder_propensity · is_disorder_promoting / DisorderPredictor.GetDisorderPropensity · .IsDisorderPromoting |
| Protein low-complexity (SEG) — from the disorder engine | predict_low_complexity_seg / DisorderPredictor.PredictLowComplexityRegions |
| Protein low-complexity (SEG) — from the motif engine | find_protein_low_complexity_regions / ProteinMotifFinder.FindLowComplexityRegions |
| N-terminal signal peptide + cleavage site | predict_signal_peptide / ProteinMotifFinder.PredictSignalPeptide |
| Transmembrane helices + membrane topology | predict_transmembrane_helices / ProteinMotifFinder.PredictTransmembraneHelices |
| Coiled-coil regions (heptad / MTIDK-style) | predict_coiled_coils / ProteinMotifFinder.PredictCoiledCoils |
| Protein domains | find_protein_domains / ProteinMotifFinder.FindDomains |
| Common protein sequence motifs | find_protein_motifs / ProteinMotifFinder.FindCommonMotifs |
| Scan a PROSITE pattern in a protein | find_motif_by_prosite / ProteinMotifFinder.FindMotifByProsite |
| Compile a PROSITE pattern → regex | prosite_to_regex / ProteinMotifFinder.ConvertPrositeToRegex |
| Kyte–Doolittle hydrophobicity profile | hydrophobicity_profile / SequenceStatistics.CalculateHydrophobicityProfile |
⚠ Envelope note. predict_disorder's per-residue / per-region confidence is DISORDER-REGION-001
(uncalibrated). See the STOP rule below. The region boundaries themselves follow the validated TOP-IDP
threshold and are fine to report.
Canonical dual-mode pipelines
Coordinates: protein positions are 0-based residue indices; regions are [start,end] inclusive per the
tool doc. Report the base explicitly.
(a) Disorder + IDRs → MoRFs (intrinsic-disorder characterization)
- [MCP]
predict_disorder(sequence, windowSize?=21, disorderThreshold?=0.542, minRegionLength?=5) → residuePredictions, disorderedRegions[{start,end,meanScore,confidence,regionType}], overallDisorderContent.
- [MCP]
predict_morfs(sequence, minLength?=10, maxLength?=25) → MoRF intervals within disordered context.
- [C# API]
DisorderPredictor.PredictDisorder(...) → .PredictMoRFs(...).
- Envelope: report region boundaries +
overallDisorderContent; treat confidence as uncalibrated (DISORDER-REGION-001). If a validated per-region call is required, use the DisorderPredictor.PredictDisorderRegions branch and STOP rather than raising the mode (below).
- Cross-check: an IDR should overlap a hydrophobicity minimum from pipeline (e) and (usually) a protein SEG low-complexity region.
(b) Membrane / secretion topology → signal peptide + TM helices
- [MCP]
predict_signal_peptide(proteinSequence, prokaryote?=false) → signal-peptide presence + cleavage site.
- [MCP]
predict_transmembrane_helices(proteinSequence, windowSize?=19, threshold?=1.6) → TM segments + in/out topology.
- [C# API]
ProteinMotifFinder.PredictSignalPeptide(...) → .PredictTransmembraneHelices(...).
- Cross-check: each TM segment should sit on a hydrophobicity peak from pipeline (e); a called signal peptide occupies the N-terminus and should not be double-counted as TM1.
(c) Structural / functional motifs → coiled-coils + domains + PROSITE
- [MCP]
predict_coiled_coils(proteinSequence, windowSize?=28, threshold?=0.5) → coiled-coil intervals + score.
- [MCP]
find_protein_domains(proteinSequence) → domain calls; find_protein_motifs(proteinSequence) → common motifs.
- [MCP] targeted PROSITE:
find_motif_by_prosite(proteinSequence, prositePattern, motifName?) → hits; prosite_to_regex(prositePattern) inspects the compiled regex first.
- [C# API]
ProteinMotifFinder.PredictCoiledCoils(...) · .FindDomains(...) · .FindCommonMotifs(...) · .FindMotifByProsite(...) · .ConvertPrositeToRegex(...).
- Cross-check: confirm a PROSITE hit's residue span against
prosite_to_regex before interpreting.
(d) Protein low-complexity (SEG) masking before motif interpretation
- [MCP]
predict_low_complexity_seg(sequence, triggerWindow?=12, triggerThreshold?=2.2, extensionThreshold?=2.5, minLength?=1) or find_protein_low_complexity_regions(proteinSequence, windowSize?=12, triggerComplexity?=2.2, extensionComplexity?=2.5) → SEG intervals.
- [C# API]
DisorderPredictor.PredictLowComplexityRegions(...) or ProteinMotifFinder.FindLowComplexityRegions(...).
- Both are protein SEG. Use as a filter: down-weight domain/PROSITE hits that fall inside SEG regions (compositionally biased, prone to spurious matches).
(e) Hydrophobicity profile (supporting evidence)
- [MCP]
hydrophobicity_profile(proteinSequence, windowSize?=9) → per-window Kyte–Doolittle score.
- [C# API]
SequenceStatistics.CalculateHydrophobicityProfile(...).
- Feeds cross-checks: peaks corroborate TM helices (b); troughs corroborate IDRs (a).
Envelope — STOP rule (guarded unit in scope)
- DISORDER-REGION-001 (
predict_disorder / DisorderPredictor.PredictDisorder, output the per-residue /
per-region confidence) is Permissive-only (blocked in Strict & Moderate; default is Moderate). No
disorder predictor publishes a calibrated confidence standard. If a task requires a validated per-region
confidence and the call would run below its MinimumMode (throws SeqeronLimitationException), STOP and
report the envelope — do not raise the mode to force output (bio-rigor rule). Named alternative: the
region boundaries are validated (TOP-IDP threshold), so report those + overallDisorderContent via
DisorderPredictor.PredictDisorderRegions; treat confidence as advisory only. The other tools in this
family (signal peptide, TM, coiled-coil, domains, motifs, SEG, hydrophobicity) are not guarded.
End-to-end grounded example — "characterize this protein"
Task. Given one protein sequence, (1) map intrinsically disordered regions, (2) test for a signal peptide,
(3) find transmembrane helices, (4) call coiled-coils + domains, (5) mask protein low-complexity, (6) profile
hydrophobicity to corroborate (1) and (3).
Tool / Method ID chain (MCP names; C# path in parentheses):
predict_disorder(sequence, minRegionLength=5) → disorderedRegions (report boundaries + overallDisorderContent; confidence advisory). (DisorderPredictor.PredictDisorder)
predict_signal_peptide(proteinSequence, prokaryote=false) → N-terminal signal + cleavage site. (ProteinMotifFinder.PredictSignalPeptide)
predict_transmembrane_helices(proteinSequence, threshold=1.6) → TM segments + topology. (ProteinMotifFinder.PredictTransmembraneHelices)
predict_coiled_coils(proteinSequence) + find_protein_domains(proteinSequence) → coiled-coil + domain calls. (ProteinMotifFinder.PredictCoiledCoils / .FindDomains)
find_protein_low_complexity_regions(proteinSequence) → SEG intervals to down-weight motif hits. (ProteinMotifFinder.FindLowComplexityRegions)
hydrophobicity_profile(proteinSequence, windowSize=9) → peaks over TM (step 3), troughs over IDRs (step 1). (SequenceStatistics.CalculateHydrophobicityProfile)
Provenance
1) predict_disorder(sequence, minRegionLength=5) → disorderedRegions [0-based, boundaries validated; confidence advisory], overallDisorderContent
2) predict_signal_peptide(proteinSequence, prokaryote=false) → signal peptide + cleavage site
3) predict_transmembrane_helices(proteinSequence, threshold=1.6) → TM segments + topology
4) predict_coiled_coils(proteinSequence); find_protein_domains(proteinSequence) → coiled-coil + domains
5) find_protein_low_complexity_regions(proteinSequence) → SEG intervals (mask/down-weight)
6) hydrophobicity_profile(proteinSequence, windowSize=9) → peaks corroborate TM (3), troughs corroborate IDRs (1)
Coordinates: 0-based residue indices; regions [start,end] inclusive. Units: normalized scores / bits / kcal-scale KD.
Envelope: DISORDER-REGION-001 — report validated boundaries + overallDisorderContent; confidence uncalibrated (STOP if a validated per-region confidence is required).
Caveat: alpha software; not for clinical use — independently validate before any decision use.
Reference