# Subpfam Functional Resolution Augmentation > Use when you have annotated genes with Pfam domains but require higher functional specificity to detect natural product sub-clusters. - Skill: `holobiomicslab/subpfam-functional-resolution-augmentation` (Agent Skill) - Install (CLI): `npx skillmds@latest add holobiomicslab/subpfam-functional-resolution-augmentation` - Raw SKILL.md: https://api.skillmd.com/api/skills/holobiomicslab/subpfam-functional-resolution-augmentation/raw - Safety review: CAUTION (external: skill-scanner FAIL, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - License: CC-BY-4.0 - Author: HolobiomicsLab (https://skillmd.com/u/holobiomicslab) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/holobiomicslab/subpfam-functional-resolution-augmentation --- # subpfam-functional-resolution-augmentation ## Summary This skill augments Pfam domain annotations with subPfam assignments to increase functional resolution when tokenizing individual genes in Biosynthetic Gene Clusters. It enables finer-grained domain-level representation suitable for detecting functionally coherent sub-clusters within larger BGCs. ## When to use Apply this skill when you have annotated genes with Pfam domains but require higher functional specificity to detect natural product sub-clusters. Use it specifically when Pfam-level annotation alone produces over-broad domain tokens that obscure functionally distinct sub-domains, or when linking detected sub-clusters to specific natural product substructures requires domain-level granularity beyond standard Pfam assignments. ## When NOT to use - Input genes have not yet been scanned for Pfam domains; perform standard Pfam annotation first. - Downstream analysis does not require sub-cluster detection or natural product substructure linking; standard Pfam tokens may be sufficient for coarser-grained clustering tasks. - Genes belong to non-biosynthetic pathways where Pfam/subPfam domain architecture is not the primary discriminator. ## Inputs - GenBank-format BGC sequence files - Gene sequences annotated with Pfam domain assignments - Pfam domain identifiers and boundaries per gene ## Outputs - Tokenized BGC representation (one token string per gene, combining Pfam domains and subPfams) - Gene-level annotation objects enriched with subPfam metadata ## How to apply After initial Pfam domain scanning of BGC genes, systematically augment each Pfam annotation with its corresponding subPfam assignment to increase resolution. Represent each gene as a token string combining both Pfam domains and their subPfam refinements. This dual-level representation preserves broad functional categories (via Pfams) while capturing fine-grained functional distinctions (via subPfams) needed for subsequent sub-cluster detection methods like PRESTO-STAT or PRESTO-TOP. The rationale is that subPfams partition Pfam domains into functionally distinct subtypes, allowing the similarity network filtering and sub-cluster detection steps to discriminate between genes with related but non-identical biosynthetic roles. ## Related tools - **iPRESTO** (Command-line tool that executes BGC tokenization using Pfam and subPfam domains as part of its integrated sub-cluster detection pipeline) - **Pfam domain scanner** (Upstream tool for initial Pfam domain assignment; provides the base annotations that are subsequently augmented with subPfams) ## Evaluation signals - Each gene in the tokenized output contains both Pfam and subPfam identifiers in its token string; verify no gene is represented by Pfam domains alone. - Token strings show increased specificity compared to Pfam-only representation; subPfam tokens refine broad Pfam categories into distinct functional subtypes. - Downstream similarity network filtering produces non-trivial redundancy reduction, indicating that subPfam augmentation enabled meaningful discrimination among BGCs. - Sub-clusters detected after augmentation show tighter functional coherence and stronger linkage to specific natural product substructures than pre-augmentation results. - Audit log or token comparison confirms subPfam assignments are non-empty and systematic across all genes; no missing or spurious subPfam entries. ## Limitations - Augmentation depends on the completeness and accuracy of the underlying Pfam domain database; genes lacking Pfam annotation cannot be augmented. - SubPfam assignments may not exist for all Pfam domains; augmentation coverage may be partial, requiring fallback to Pfam-only representation for unaugmented domains. - Increased dimensionality from dual-level annotation may inflate computational cost in large-scale BGC screening; trade-off between resolution and scalability should be considered. - The quality of sub-cluster detection and natural product linkage depends critically on downstream filtering and statistical methods (PRESTO-STAT, PRESTO-TOP); augmentation alone does not guarantee valid or interpretable results. ## Evidence - [other] Each gene in a BGC is tokenised by representing it as a combination of its Pfam domains, with subPfams employed to increase resolution of the domain annotation.: "Each gene in a BGC is tokenised by representing it as a combination of its Pfam domains, with subPfams employed to increase resolution of the domain annotation." - [intro] BGCs are tokenised by representing each gene as a combination of its Pfam domains, where subPfams are used to increase resolution: "BGCs are tokenised by representing each gene as a combination of its Pfam domains, where subPfams are used to increase resolution" - [other] Represent each gene as a token string combining its Pfam domains and subPfams.: "Represent each gene as a token string combining its Pfam domains and subPfams." - [intro] The sub-clusters found with iPRESTO can then be linked to Natural Product substructures: "The sub-clusters found with iPRESTO can then be linked to Natural Product substructures" - [intro] Tokenise BGCs using Pfam domains and subPfams: "iPRESTO detects gene sub-clusters in Biosynthetic Gene Clusters by tokenizing BGCs using Pfam domains and subPfams"