GLKB iconAPI Docs

Variant & GWAS Interpretation

Prioritize candidate causal genes at GWAS loci with literature evidence.

When to use

You have a list of variants from a GWAS, whole-exome/genome sequencing, or fine-mapping analysis and need to prioritize which genes are most likely functional and disease-relevant before investing in experimental follow-up.

Example: GWAS fine-mapping prioritization

A T2D GWAS identifies 200 fine-mapped loci. For each credible set gene, the question is: does the literature support a functional mechanism linking this gene to beta cell biology or glucose homeostasis? Manual curation at 200 genes takes weeks; the batch endpoint returns a prioritized list with citations in under an hour.

Prompt template

python
GWAS_PROMPT = """Evaluate the gene {item} as a candidate causal gene at a Type 2 diabetes GWAS locus.

Search GLKB and PubMed for:
1. Any published functional evidence linking {item} to beta cell biology, insulin
   secretion, insulin resistance, or glucose homeostasis
2. Human genetic evidence beyond GWAS: rare coding variants, eQTL colocalisation,
   Mendelian phenotypes
3. Mouse or cell-line KO/KD phenotypes relevant to diabetes
4. Expression: is {item} expressed in human islets, liver, or muscle?

### {item}
- **Functional evidence:** [1-2 sentences or "none found"]
- **Genetic evidence beyond GWAS:** [evidence or "none found"]
- **Model organism phenotype:** [phenotype or "none found"]
- **Islet expression:** [Yes / No / Unknown]
- **Priority:** [High / Medium / Low]
- **PMIDs:** [up to 3]"""

Code

python
import json, re, requests, pandas as pd, os

def extract_field(text, field):
	m = re.search(rf"\*\*{field}:\*\*\s*(.+?)(?=\n-|\Z)", text, re.S)
	return m.group(1).strip() if m else ""

locus_genes = ["TCF7L2", "KCNJ11", "SLC30A8", "CDKAL1", ...]  # 200 GWAS genes
records = []

with requests.post(
	"https://api.glkb.dev/apps/glkb/users/me/batch",  # <!-- DEV: confirm URL -->
	headers={"Authorization": f"Bearer {os.environ['GLKB_API_KEY']}"},
	json={"items": locus_genes, "prompt_template": GWAS_PROMPT, "concurrency": 10},
	stream=True,
	timeout=7200,
) as r:
	r.raise_for_status()
	for line in r.iter_lines():
		if line and line.startswith(b"data:"):
			ev = json.loads(line[5:])
			if ev.get("status") == "done":
				records.append({
					"gene":     ev["item"],
					"priority": extract_field(ev["response"], "Priority"),
					"summary":  extract_field(ev["response"], "Functional evidence"),
					"pmids":    extract_field(ev["response"], "PMIDs"),
				})

df = pd.DataFrame(records)
df[df["priority"] == "High"].to_csv("high_priority_loci.csv", index=False)

Sample output

gene,     priority,  summary,                                          pmids
TCF7L2,   High,      Regulates Wnt signaling in beta cell survival,    12345678 23456789
KCNJ11,   High,      Encodes Kir6.2 subunit of KATP channel in islets, 34567890
CDKAL1,   Medium,    tRNA modification role; indirect islet evidence,  none found

A T2D GWAS identifies 200 fine-mapped loci. For each credible set gene, the question is: does the literature support a functional mechanism linking this gene to beta cell biology or glucose homeostasis? Manual curation at 200 genes takes weeks; the batch endpoint returns a prioritized list with citations in under an hour.

Extending to rare variants

Replace locus gene names with GENE:p.VAR strings and ask the agent to retrieve functional domain context, known pathogenic variants at the same residue, and ClinVar classifications. The same prompt structure applies.