Colorectal cancer: research and therapeutic landscape

12 genes recurrently implicated in colorectal cancer, triaged for whether anything can actually be aimed at them. The zeros are the informative rows.

Of 12 recurrently implicated genes, 8 carry any drug at all. That is a statement about these 12 gene targets, not about the disease: Colorectal cancer does have labelled therapy — 15 drugs, listed in the next section.

Dataset evidence counts studies in the 2072-study ranked set whose title or abstract names the gene; the bar is scaled to KRAS. Sources: Open Targets Platform and NCBI GEO, retrieved 2026-09-30. Clinical stage is the highest reached for any indication — see the per-drug indications noted in each row. Target-prioritisation reference only; not clinical or prescribing guidance.

15 drugs carry an FDA label naming colorectal cancer: Adagrasib, Bevacizumab, Cetuximab, Encorafenib, Fruquintinib, Ipilimumab, Nivolumab, Panitumumab, Pembrolizumab, Ramucirumab, Regorafenib, Sotorasib, Trifluridine And Tipiracil, Tucatinib, Ziv-Aflibercept. Separately, 29 of the drugs returned for the genes in the table above are approved only for other diseases and reach colorectal cancer through trials, not through their labels.

Every label that names colorectal cancer

113 labels carry a positive indication naming the disease under indications_and_usage:"colorectal cancer" OR indications_and_usage:"colon cancer" OR indications_and_usage:"rectal cancer" OR indications_and_usage:"carcinoma of the colon or rectum" OR indications_and_usage:"colorectal carcinoma"; they collapse to 22 distinct molecules once salt forms, biosimilars and co-formulated hyaluronidase are merged. Supportive agents (Leucovorin Calcium, Levoleucovorin, Levoleucovorin Calcium) are given in the disease but do not act on it. Source: openFDA drug label API (api.fda.gov/drug/label.json), retrieved 2026-09-18. Labels change; this reflects the current label text, not the approval history. Not prescribing guidance.

Labels that name colorectal cancer without indicating for it

A label can name a disease under Limitations of Use, in a contraindication or as a risk. These were read and are not counted above.

CEA is the busiest cell-therapy target among those this briefing searched in colorectal cancer: 18 registered trials naming it, 8 still active, 2 withdrawn before enrolling anyone. It has no gene entry of its own and is reached through CEACAM5: the surface glycoprotein CEACAM5 encodes, the CEA of the blood test.

10 of 18 shown, most recently active first. Other targets searched: NKG2D (7), PD-1 (6), EGFR (3), MUC1 (3), GUCY2C (1), HER2 (1), TROP2 (1), LGR5 (1). Source: ClinicalTrials.gov API v2, retrieved 2026-09-18. A trial counts here when its registered conditions name colorectal cancer or a tumour-agnostic basket, its title names an administered cell product, and the title names this target: a HER2 CAR-T trial that adds a PD-1 antibody is a HER2 trial. A trial stating only a product code is missed.

Every term below is a share of the colorectal cancer literature, not a count, because the field itself grew: 2015–2018 (n=6,000) against 2021–2025 (n=6,000). A topic whose papers doubled while the field doubled has not risen.

These are sample shares. The two windows hold 32,289 and 48,775 papers; MeSH terms were read from an evenly spaced sample of 6,000 and 6,000 of them, taken across the whole window rather than from its most recent papers. Percentages carry sampling error of roughly a percentage point and small differences between two terms are not meaningful.

Rising, with a real baseline

Cooling

Biggest topics now

Publication mix

Query: Colorectal Neoplasms[MeSH Major Topic] NOT ("Anus Neoplasms"[MeSH] OR "Neuroendocrine Tumors"[MeSH] OR "Appendiceal Neoplasms"[MeSH]). Terms need at least 12 papers in the recent window and, in the rising table, at least 5 in the earlier one — a term going from 1 paper to 12 is a large fold change and no evidence of anything. Ubiquitous descriptors (Humans, Animals, age bands) and the disease’s own term are dropped. Windows are read whole where they fit and sampled where they do not; the totals and the sample sizes are both in the JSON beside this page. Source: PubMed MeSH, retrieved 2026-09-12.

What the public sources count, and how closely each category matches this disease. The same figures for every disease on the site are on one table.

Years of life lost (proxy)

4.3 years per case, 681,530 a year, from survival and age at diagnosis rather than from a death count. derived proxy Not the WHO/GBD measure, which uses age at death and a life table; comparable across these briefings, not with published YLL figures. What this costs to fund is in Funding.

NIH obligations naming colorectal cancer, after removing the 9,685 records across all years that matched the search but are about something else. That filter is not cosmetic: without it this section reports another field’s funding as this disease’s.

NIH obligations by fiscal year

Where FY2025 money went

Text search colorectal cancer over project title, terms and abstract, per fiscal year. Awards are individual funding actions; distinct projects collapse them by core project number, so a multi-year grant counts once. The most recent year may be incomplete. Source: NIH RePORTER, retrieved 2026-09-19.

What that buys

Against 681,530 years of life lost a year, FY2025 obligations are $311 per life-year — $1,333 per case. The estimate and its assumptions are in Disease burden.

Who funds it, FY2025

The total says how much. These say who from, through what kind of award, and to whom — which a dollar figure cannot.

5,232 human GEO series match colorectal cancer. Keyword relevance cannot tell a 2,503-sample patient cohort from a six-well cell-line experiment, so this ranking scores four things GEO does not expose: how often the accession is named in full-text papers, field-normalised citation impact, clinical annotation, and cohort type.

Composition of the 2072-study ranked set

Sources: NCBI GEO + Europe PMC + iCite, retrieved 2026-09-19. SubSeries are collapsed to one row per study by linked PMID. Of 5,101 series retrieved, 96 were dropped by the profile’s exclusion rules and 2,473 named the disease only in passing. The ranked set is a relevance-ranked sample, not a census, so the composition figures describe the sample only.

Each card below is a disagreement between two of the tables on this page — a drug that exists but was never tried here, a target that looks tested and was not, a gene with no drug whose product has a busy clinical programme. They are computed from the same facts as the tables, so they cannot contradict them.

This page is written to be quoted correctly by a model as much as read by a person. The same facts are available as JSON at the sibling URL below, which is what to cite. The cross-disease layer — burden, genes, drugs and antigens across every briefing — is described at /disease/api.json.

{
  "disease": "Colorectal cancer",
  "mesh": "Colorectal Neoplasms",
  "facts": "https://usebiotransfer.org/disease/colorectal-cancer.json",
  "methods": "https://usebiotransfer.org/methods/",
  "all_diseases": "https://usebiotransfer.org/disease/api.json",
  "note": "Every figure on the page is in that JSON with its source and retrieval date. Quote from it rather than from the HTML."
}