ScilifelabDataCentre/scilifelab-ai-agent-mcp-workshop-2026-11-02 / Section_1_LangGraph/utils/codeact_skills/skills/biology
biology
VerifiedProteins, drug targets and peptides. Look up protein function, location and diseases (UniProt), a drug's targets, mechanism of action and indications (ChEMBL), protein interaction partners (STRING), fetch protein sequences, compute sequence properties and align sequences (Biopython). Use for any question about targets, mechanisms, protein biology or protein/peptide sequences.
Biology (targets, proteins, sequences)
This skill wraps the same functions as the ReAct agent's biology tools (utils/tools/biology.py), so both agents give consistent answers. Use the ready-made functions first; use the recipes below for batches, comparisons, full sequences and alignments.
1. Ready-made functions
from utils.tools.biology import get_protein_info, get_drug_targets, get_protein_interactions, analyze_protein_sequence
get_protein_info.invoke({"gene_symbol": "ABL1"}) # organism: "human" (default), "mouse" or "rat"
# -> {'uniprot_accession', 'protein_name', 'length_aa', 'function', 'subcellular_location', 'associated_diseases'}
get_drug_targets.invoke({"drug_name": "Imatinib"}) # generic or brand name
# -> {'chembl_id', 'preferred_name', 'max_clinical_phase',
# 'mechanisms': [{'mechanism_of_action', 'action_type', 'target': {'name', 'type', 'gene_symbols'}}],
# 'indications': [{'disease', 'max_phase'}]}
get_protein_interactions.invoke({"gene_symbol": "EGFR", "limit": 10})
# -> {'interaction_partners': [{'partner', 'confidence', 'experimental_evidence', 'curated_databases'}]}
analyze_protein_sequence.invoke({"sequence": "HAEGTFTSDVSSYLEGQAAKEFIAWLVKGRG"})
# -> {'length_aa', 'molecular_weight_da', 'isoelectric_point', 'net_charge_at_pH_7.4', 'gravy',
# 'instability_index', 'predicted_stable_in_vitro', ...}
Always check for an 'error' key. Use official gene symbols (ABL1, not "BCR-ABL" or "c-Abl"). If mechanisms has no annotation it is a string, not a list.
2. Targets of several drugs → shared and unique targets
import pandas as pd
def target_genes(drug: str) -> set[str]:
res = get_drug_targets.invoke({"drug_name": drug})
if "error" in res or not isinstance(res["mechanisms"], list):
print(f"{drug}: {res.get('error', 'no mechanism annotated')}")
return set()
return {gene for m in res["mechanisms"] for gene in m["target"].get("gene_symbols", [])}
targets = {drug: target_genes(drug) for drug in ["Imatinib", "Dasatinib", "Nilotinib"]}
shared = set.intersection(*targets.values())
print("Shared targets:", shared)
pd.DataFrame({drug: {g: g in genes for g in set().union(*targets.values())} for drug, genes in targets.items()})
3. Fetch a full protein sequence (UniProt)
get_protein_info does not return the sequence (to keep the LLM context small). Fetch it when you need to compute on it, and do not print whole sequences: print the length and a short preview.
import requests
def fetch_sequence(uniprot_accession: str) -> str:
response = requests.get(f"https://rest.uniprot.org/uniprotkb/{uniprot_accession}.fasta", timeout=30)
response.raise_for_status()
return "".join(response.text.splitlines()[1:]) # drop the FASTA header line
accession = get_protein_info.invoke({"gene_symbol": "EGFR"})["uniprot_accession"]
egfr = fetch_sequence(accession)
print(len(egfr), egfr[:60] + "...")
4. Sequence properties at other conditions
analyze_protein_sequence reports charge at pH 7.4. For other questions use Biopython directly:
from Bio.SeqUtils.ProtParam import ProteinAnalysis
analysis = ProteinAnalysis("HAEGTFTSDVSSYLEGQAAKEFIAWLVKGRG")
{pH: round(analysis.charge_at_pH(pH), 2) for pH in [2.0, 5.0, 7.4, 9.0]} # e.g. stomach vs blood
5. Align two sequences (identity, differences)
from Bio import Align
from Bio.Align import substitution_matrices
def align(seq_a: str, seq_b: str):
aligner = Align.PairwiseAligner(
mode="global",
substitution_matrix=substitution_matrices.load("BLOSUM62"),
open_gap_score=-10, extend_gap_score=-0.5,
)
alignment = aligner.align(seq_a, seq_b)[0]
counts = alignment.counts()
identity = counts.identities / min(len(seq_a), len(seq_b))
return alignment, round(100 * identity, 1)
alignment, identity_pct = align("HAEGTFTSDVSSYLEGQAAKEFIAWLVKGRG", "HGEGTFTSDLSKQMEEEAVRLFIEWLKNGGPSSGAPPPS")
print(alignment)
print(f"Identity: {identity_pct}% of the shorter sequence")
Positions where the sequences differ often explain differences in stability or activity (e.g. residue 2 of GLP-1 analogues and DPP-4 cleavage).
6. Compare interaction networks
def partners(gene: str, n: int = 50) -> dict[str, float]:
res = get_protein_interactions.invoke({"gene_symbol": gene, "limit": n})
return {p["partner"]: p["confidence"] for p in res.get("interaction_partners", [])}
egfr, erbb2 = partners("EGFR"), partners("ERBB2")
common = sorted(egfr.keys() & erbb2.keys(), key=lambda g: -min(egfr[g], erbb2[g]))
print(f"{len(common)} common partners:", common[:15])
STRING confidence: ≥ 0.7 high, ≥ 0.9 highest. experimental_evidence shows how much of it comes from lab experiments.
Pitfalls
- Gene symbols are case-sensitive per organism: human
EGFR, mouseEgfr. - ChEMBL mechanisms describe the intended targets of approved/clinical drugs, not every off-target.
- Sequence-based predictions (instability index, GRAVY) ignore modifications (amidation, lipidation, PEGylation) and in vivo processes such as enzymatic cleavage or renal clearance. Back up claims about in vivo behaviour with the
literatureskill.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |