From mims-harvard-tooluniverse
Performs cross-species gene and sequence comparisons, ortholog identification, and evolutionary conservation analysis using Ensembl Compara, NCBI, UniProt, and Python tools.
npx claudepluginhub joshuarweaver/cascade-data-analytics --plugin mims-harvard-tooluniverseThis skill uses the workspace's default tool permissions.
Cross-species gene comparison, ortholog identification, sequence retrieval, and functional conservation analysis integrating Ensembl Compara, NCBI, UniProt, OLS, Monarch, and OpenTargets.
Conducts multi-round deep research on GitHub repos via API and web searches, generating markdown reports with executive summaries, timelines, metrics, and Mermaid diagrams.
Dynamically discovers and combines enabled skills into cohesive, unexpected delightful experiences like interactive HTML or themed artifacts. Activates on 'surprise me', inspiration, or boredom cues.
Generates images from structured JSON prompts via Python script execution. Supports reference images and aspect ratios for characters, scenes, products, visuals.
Cross-species gene comparison, ortholog identification, sequence retrieval, and functional conservation analysis integrating Ensembl Compara, NCBI, UniProt, OLS, Monarch, and OpenTargets.
When uncertain about any scientific fact, SEARCH databases first (PubMed, UniProt, ChEMBL, ClinVar, etc.) rather than reasoning from memory. A database-verified answer is always more reliable than a guess.
When analysis requires computation (statistics, data processing, scoring, enrichment), write and run Python code via Bash. Don't describe what you would do — execute it and report actual results. Use ToolUniverse tools to retrieve data, then Python (pandas, scipy, statsmodels, matplotlib) to analyze it.
Triggers:
Use Cases:
Understanding conservation requires distinguishing between types of evolutionary patterns and what they imply about function.
High conservation signals functional constraint. When a gene is maintained as a 1:1 ortholog from yeast to humans, purifying selection has prevented sequence divergence — the gene's function is essential and cannot be easily altered. Highly conserved positions within a protein sequence (high PhastCons scores > 0.8, or GERP RS > 4) are under strong constraint; mutations at these positions are disproportionately pathogenic. For non-coding regions, conservation in mammals at PhastCons > 0.5 suggests a candidate regulatory element.
Low conservation in one lineage has two possible explanations: relaxed selection or positive selection. Use the dN/dS ratio (nonsynonymous to synonymous substitution rate) to distinguish them. A dN/dS ratio near 1 suggests neutral evolution — the gene is no longer under purifying selection (relaxed constraint, possibly reflecting loss of function in that lineage). A dN/dS ratio > 1 indicates positive selection — the gene is diverging faster than neutral expectation, often because it is adapting to a new environment or function. A dN/dS ratio << 1 is the signature of purifying selection (functional constraint). When a vertebrate gene shows high divergence in a specific branch of the tree, ask which explanation applies before concluding that function is lost.
Ortholog relationship type shapes interpretation. A 1:1 ortholog (one gene in human, one in mouse) is the highest-confidence functional equivalent — it has not been duplicated in either lineage, so it most likely performs the same ancestral role. A 1:many relationship (one gene in human, multiple in mouse) means the target species has duplicated the gene; the copies may have subfunctionalized (each copy performs a subset of the original roles) or neofunctionalized (one copy gained a new role). Do not assume both copies retain full ancestral function. A many:many relationship reflects complex duplication history in both species and requires analyzing each paralog pair individually.
Conservation depth predicts essentiality. A gene conserved across all vertebrates suggests a fundamental cellular process. A gene conserved only in mammals suggests a more specialized vertebrate innovation. A gene present only in primates or only in humans is likely a recent evolutionary acquisition, possibly involved in human-specific biology but often lacking the depth of functional characterization available for deeply conserved genes.
Absence of an ortholog is a finding, not an error. Lineage-specific genes exist and are biologically meaningful. Before concluding a gene is lineage-specific, check: (1) whether BLAST with relaxed thresholds finds distant homologs, (2) whether a highly divergent ortholog exists that Ensembl Compara missed, and (3) whether the gene belongs to a rapidly evolving family (immune genes, olfactory receptors, reproductive proteins) where turnover is expected.
Input (gene symbol/ID + reference species)
|
v
Phase 1: Gene Identification & Validation
|
v
Phase 2: Ortholog Discovery (Ensembl Compara + OpenTargets)
|
v
Phase 3: Sequence Retrieval (NCBI + Ensembl)
|
v
Phase 4: Functional Annotation Comparison (UniProt + OLS GO terms)
|
v
Phase 5: Cross-Species Phenotype Bridging (Monarch)
|
v
Phase 6: Gene Tree & Evolutionary Context (Ensembl Compara)
|
v
Report: Conservation summary, ortholog evidence, functional comparison, phenotype bridging
ensembl_lookup_gene takes gene_id (symbol or Ensembl ID). The species parameter is REQUIRED when using gene symbols (e.g., species="homo_sapiens"); omitting it causes errors. Extract the Ensembl gene ID, description, biotype, and chromosomal coordinates for downstream queries. For non-human references, adjust species accordingly (e.g., "mus_musculus", "danio_rerio").
EnsemblCompara_get_orthologues is the primary tool. It takes gene (symbol or Ensembl ID), species (source species, default "human"), and optionally target_species (e.g., "mouse", "zebrafish") or target_taxon (NCBI taxon ID). Omit target_species to get all orthologs across the tree; filter client-side for specific species. It returns homology type (one2one, one2many, many2many) and the taxonomy divergence level for each ortholog.
ensembl_get_homology is the alternative when you need sequence-level data alongside the ortholog mapping. Use sequence="protein" and aligned=true for aligned sequence comparison across species.
OpenTargets_get_target_homologues_by_ensemblID (takes ensemblId) provides supplementary ortholog data from OpenTargets, which can add druggability context and cross-reference with model organism phenotype data.
Reasoning: Prioritize 1:1 orthologs as high-confidence functional equivalents. For 1:many cases, report all copies and flag the need for paralog-specific functional analysis. If no Ensembl Compara entry exists, try BLAST as a last resort (note: BLAST protein search against swissprot is slow, 5-30 minutes; against nr may take longer).
Key model organisms to check: mouse (taxon 10090), rat (10116), zebrafish (7955), fruit fly (7227), C. elegans (6239), S. cerevisiae (4932).
Use NCBI_search_nucleotide (takes organism as full name, e.g., "Homo sapiens"; gene; seq_type = "mRNA") to find sequence records, then NCBI_fetch_accessions to convert UIDs to accession numbers, then NCBI_get_sequence to retrieve FASTA data. Prefer RefSeq (NM_* for mRNA, NP_* for protein) over other accessions for canonical sequence.
When aligned sequences are needed directly, ensembl_get_homology with sequence="cdna" or sequence="protein" is faster than running BLAST. Use BLAST only when Ensembl Compara does not find orthologs.
UniProt_search takes a query in UniProt syntax (e.g., "gene:TP53 AND organism_id:9606 AND reviewed:true") and fields to retrieve specific annotation columns including GO terms. Use reviewed:true to restrict to Swiss-Prot curated entries.
UniProt_get_function_by_accession takes a UniProt accession and returns a list of function description strings (not a dict).
For each species being compared, retrieve GO terms and group them by Biological Process (BP), Molecular Function (MF), and Cellular Component (CC). Shared GO terms indicate conserved function; terms present in human but absent in the ortholog may reflect annotation bias (less-studied organisms have fewer GO annotations) rather than true functional divergence. Focus conservation claims on shared terms.
Reasoning about annotation gaps: If a mouse ortholog lacks a GO term present in the human protein, consider that this may reflect incomplete annotation of the mouse gene rather than functional divergence. The inverse — a GO term in mouse that is absent in human — is less common but can indicate diverged or acquired function.
Monarch_search_gene (takes query as gene symbol) returns gene CURIEs needed for Monarch queries. Monarch_get_gene_phenotypes and Monarch_get_gene_diseases take a gene CURIE (e.g., "HGNC:11998") and return phenotype/disease associations spanning multiple species.
Phenotype ontologies by species: Human = HP (HPO), Mouse = MP (Mammalian Phenotype), Zebrafish = ZP, Fly = FBcv. Monarch integrates across species; compare phenotype themes (e.g., "tumor susceptibility" in human and "increased tumor incidence" in mouse) rather than requiring exact term matches.
Reasoning for model organism selection: A mouse ortholog that has a 1:1 relationship AND shows phenotypes in Monarch that recapitulate the human disease is a strong disease model candidate. If the mouse phenotype diverges significantly from the human disease phenotype, this is worth flagging — it could indicate species-specific function or a limitation of the model.
EnsemblCompara_get_gene_tree (takes gene, species) returns the gene tree members, species distribution, and speciation vs. duplication events. EnsemblCompara_get_paralogues returns all paralogs in the source species.
From the gene tree, assess: (1) how many species contain a member of this gene family; (2) when gene duplication events occurred (ancient vs. recent); (3) whether the gene family expanded in particular lineages. A gene present in a single copy across all vertebrates (deep conservation, no duplication) is likely under strong selective constraint.
When interpreting the assembled evidence, work through these questions:
ensembl_get_homology, then OpenTargets_get_target_homologues_by_ensemblID, then BLAST as last resortensembl_get_homology with sequence="cdna" as alternative to NCBIMonarchV3_get_associations with category="biolink:GeneToPhenotypicFeatureAssociation" as alternative