Computational prioritization of cysteine-rich secreted-like peptide candidates from psilocybin-producing fungal proteomes

Fungal proteomes encode many short proteins and peptide-like sequences whose biological roles are incompletely understood. In psilocybin-producing fungal lineages, public attention often centers on secondary metabolites, but predicted proteomes provide a separate computational substrate for asking whether short cysteine-rich, secreted-like peptide candidates are present across available assemblies.

This study frames those sequences as a reproducible sequence-space survey. It does not infer therapeutic relevance, biological activity, secretion, or risk profile from sequence alone. Instead, the goal is to create a transparent family-level map that can support future human scientific review.

The acquisition phase selected four assemblies that exposed the required NCBI protein FASTA, genomic GFF, and assembly report files: Psilocybe cubensis, Psilocybe cyanescens, Panaeolus cyanescens, and Gymnopilus dilepis. Psilocybe azurescens and Pluteus salicinus were excluded because the acquisition manifest did not find a required protein FASTA plus genomic GFF package.

Reference-space comparison used the acquired local reference archive including DBAASP, DRAMP, APD, MEROPS, and PeptideAtlas where available. DBAASP was integrated as a local-first snapshot from dbaasp.org for nearest-neighbor retrieval, reference-space analysis, and annotation context only.

The campaign retained 470 mature peptide candidates after filtering and global deduplication. Candidate IDs, precursor sequences, mature sequences, species, source assembly, source database, computed features, and provenance were preserved in native Protean candidate objects.

The public study reports the existing candidate and family outputs. It does not modify the top-25 ranking, scoring stack, weights, validators, frontier behavior, or database-comparison effects.

The DBAASP snapshot contained 25,069 raw records, 20,582 canonical normalized peptide rows, 16,989 unique peptide sequences, and 171 hydrated detail records. The snapshot was local-first and introduced no live scoring dependency.

Candidate-level DBAASP nearest-neighbor comparison found 0 exact matches, 0 matches at or above 90 percent identity, 4 matches at or above 80 percent identity, 26 at or above 70 percent identity, and 441 at or above 50 percent identity. The median nearest-neighbor identity was 0.5625 and the maximum was 0.875. DBAASP activity, toxicity, hemolysis, structure, and literature annotations, when present, describe nearest neighbors only and are not labels for Protean candidates.

The results suggest that psilocybin-producing fungal proteomes contain a broad computational landscape of short cysteine-rich, secreted-like peptide candidates. Family collapse matters because the original count of 470 candidates includes both singleton sequence hypotheses and larger repeated architectures.

The recurring motifs should be interpreted as sequence architectures. Some motifs recur across multiple families, but recurrence alone does not establish homology, secretion, activity, or ecological function. The strongest contribution of the campaign is a reproducible, artifact-backed prioritization surface for future review.

The public study is backed by campaign manifests, Phase II validation artifacts, DBAASP snapshot verification, neighbor JSONL outputs, figure-generation scripts, and a public claims audit. Figure data are generated from the existing campaign outputs rather than synthetic or decorative data.

The publication build preserves the original discovery outputs. The website layer converts existing artifacts into public-safe narrative, figures, tables, metadata, and dynamic social images.

Protean's Psilocybe Secretome Campaign moved from 470 computational mature peptide candidates to a family-level map of 346 cysteine-rich, secreted-like candidate families. The family structure, recurring motifs, cross-species families, and local reference-space distances provide a more defensible scientific view than candidate count alone.

The study's strongest claim is deliberately bounded: this is a reproducible computational survey and prioritization resource. It is not a validation study, not an activity claim, and not a therapeutic claim.