NCI Cancer Model Development for the Human Cancer Model Initiative

Written and checked by a reviewer, not generated from metadata.

1. 1. How faithfully do patient-derived models retain the genomics of their source tumor?

   Needs approved accessn ≈ 721HUMAN

   Models and their source tumors are both characterized, with near-complete whole genome coverage, which makes fidelity measurable rather than assumed. The answer determines how much any model-based result should be trusted.

   Statistical note: Verify model and source tumor pairing per case; it is not uniform.

   NeedsWhole genome sequencingBulk RNA sequencingModel system

2. 2. Can a usable model system be found for a rare cancer that currently has none?

   Answerable as shippedn ≈ 805HUMAN

   Coverage spans many tumor types including rare ones. For a researcher working on a cancer with no established cell line, this is often the only route to any model at all, and it is the cohort's most practically valuable use.

   Statistical note: Per-tumor-type counts are small; this is model discovery, not a powered comparison.

   NeedsModel systemWhole genome sequencing

3. 3. Does prior treatment of the donor shape the derived model's biology?

   Needs approved accessn ≈ 716HUMAN

   Prior treatment is informative for 88.9% of donors and treatment given is recorded for all, so models derived from treated and untreated patients can be compared - a question with direct bearing on how model-based drug screens should be interpreted.

   NeedsWhole genome sequencingBulk RNA sequencingTreatment recordsdiagnoses.prior_treatmentdiagnoses.treatments.treatment_type

Written by a reviewer against the source, not inferred from what is missing.

Articles that analyzed these data, counted separately from articles that only cited the paper. Only the first counts as reuse.

Analyses someone ran on these data end to end, each with a receipt recording when it ran, against which packages and for how long.

Starter code generated from this record’s own identifiers, then the access steps.

import json, requests

PROJECT = "HCMI-CMDC"
filters = {"op": "and", "content": [
    {"op": "in", "content": {"field": "cases.project.project_id", "value": [PROJECT]}},
    {"op": "in", "content": {"field": "access", "value": ["open"]}},
]}
params = {"filters": json.dumps(filters),
          "fields": "file_id,file_name,data_category,data_type,file_size",
          "size": "10000", "format": "JSON"}
files = requests.get("https://api.gdc.cancer.gov/files", params=params, timeout=60).json()["data"]["hits"]
print(len(files), "open files")

# Write a manifest and fetch with the GDC client (https://gdc.cancer.gov/access-data/gdc-data-transfer-tool)
with open("manifest.txt", "w") as fh:
    fh.write("id\n" + "\n".join(f["file_id"] for f in files))
# $ gdc-client download -m manifest.txt

import json, requests

filters = {"op": "in", "content": {"field": "project.project_id", "value": ["HCMI-CMDC"]}}
params = {"filters": json.dumps(filters), "size": "5000", "format": "TSV",
          "fields": ",".join([
              "submitter_id", "demographic.vital_status", "demographic.days_to_death",
              "diagnoses.days_to_last_follow_up", "diagnoses.ajcc_pathologic_stage",
              "diagnoses.treatments.treatment_type", "diagnoses.treatments.treatment_outcome"])}
tsv = requests.get("https://api.gdc.cancer.gov/cases", params=params, timeout=60).text
open("HCMI-CMDC_clinical.tsv", "w").write(tsv)

from idc_index import IDCClient

COLLECTION = "hcmi_cmdc"
client = IDCClient()

series = client.index[client.index["collection_id"] == COLLECTION]   # one row per DICOM series
print(len(series), "series,", series["PatientID"].nunique(), "patients,",
      round(series["series_size_MB"].sum() / 1024, 1), "GB")
print(series["Modality"].value_counts().to_dict())

# Everything above is an index lookup and costs nothing. Uncomment to pull the pixels;
# the client refuses up front if the disk cannot hold them.
# client.download_from_selection(collection_id=COLLECTION, downloadDir="./hcmi_cmdc")

curl -s https://cd2s.vercel.app/data/datasets/gdc-hcmi-cmdc.json | python3 -c '
import json, sys
r = json.load(sys.stdin)
for l in r["limitations"]:
    print(l["severity"].upper(), "-", l["statement"][:120])
print("survival endpoint:", r["longitudinal"]["has_survival_endpoint"], "| treatment response:", r["longitudinal"]["has_treatment_response"])'

Access steps

1. 1

   Search the model catalog for your tumor typeFor people1 hour

   https://portal.gdc.cancer.gov/projects/HCMI-CMDC
2. 2

   Check model availability through the HCMI Searchable CatalogFor people1 hour

   The data describe models; obtaining the physical model is a separate step.

   https://ocg.cancer.gov/programs/HCMI
3. 3

   Submit a dbGaP data access request for sequence dataFor peopledays to a few weeks

   Requires: eRA Commons account, institutional signing official approval

   https://gdc.cancer.gov/access-data/obtaining-access-controlled-data
4. 4

   Query the project from code, with no credentialsFor agents

   The file index is public even where files are not, so an agent can size the cohort before anyone requests access.

   curl -s 'https://api.gdc.cancer.gov/files' --get \
     --data-urlencode 'filters={"op":"in","content":{"field":"cases.project.project_id","value":["HCMI-CMDC"]}}' \
     --data-urlencode 'fields=file_id,file_name,data_type' \
     --data-urlencode 'size=10000' --data-urlencode 'format=TSV'

   https://docs.gdc.cancer.gov/API/Users_Guide/Getting_Started/

   Generated from repository policy. Check the linked policy before downloading.

5. 5

   Pass the token when the agent needs controlled filesFor agents

   Without a token, an API query may return only the open subset. Compare returned counts with the record before treating them as the whole cohort.

   Requires: GDC authentication token

   curl -H "X-Auth-Token: $GDC_TOKEN" https://api.gdc.cancer.gov/data/<file_id>

   https://docs.gdc.cancer.gov/API/Users_Guide/Getting_Started/

   Generated from repository policy. Check the linked policy before downloading.

All of it, and more, is in this record’s JSON.

Same disease area, favoring cohorts that add a measurement this one lacks.