What does this atlas show?
Transcription factors (TFs) are proteins that help control which genes a cell uses. This atlas shows which human TFs have reports of small chemical compounds that attach directly to them. These compounds are called binders or ligands.
To count, a compound must have an experimental measurement showing sufficiently strong binding, also called affinity. We use a cutoff below one micromolar (1 µM, or 1,000 nanomolar [nM]), measured as a binding constant called Kd or Ki. Lower values generally indicate tighter binding under the test conditions. Exactly 1 µM does not qualify.
Binding alone does not show that a compound changes the TF’s function, works in cells, is safe or could become a drug.
Which TFs are included?
The starting list is the Lambert catalogue of human TFs, version 1.01. Classes group TFs by features of the part of the protein that recognizes DNA. TBXT is also searchable by its former name, T. We added TOX to retain evidence from our earlier review, although the catalogue does not classify it as a TF that recognizes specific DNA sequences.
Nuclear receptors are a TF class that includes receptors for hormones and other small molecules. We search this class alongside all others. The toggle changes what you see, including both the TF and class counts; it does not change what we search. Counts refer to distinct genes, not the number of compounds or protein variants. The class total leaves out records whose class is unknown.
Source catalogue of human TFs ↗What evidence counts?
The automated check looks for the compound’s identity, the human TF tested, the experimental method, the numerical binding result and the original source. It also checks whether researchers tested the whole protein, a fragment or an altered version. Binding measured for one TF or compound is not assumed to apply to a related one. Results limited to an altered protein are labeled.
A cell response or a decrease in protein amount does not, by itself, measure direct binding. Neither do computer predictions or changes in protein stability when heated. Measurements that depend on an additional binding partner are not treated as measurements of the compound and TF alone. Reports of activity without a qualifying binding measurement appear separately as candidates.
We include naturally occurring and laboratory-made small molecules. We exclude peptides and larger biological treatments, compounds that act only through another target or DNA, and compounds that form a covalent chemical bond with the TF. If a patent reports a range, the entire range must be below the cutoff. Unclear results are flagged rather than assumed to qualify; unresolved details in older records remain visible.
How certain are the findings?
Reversible means a compound can attach to the TF and later detach. For many entries, this is inferred from how binding was measured or from the observed molecular structure. Those observations do not necessarily demonstrate detachment. The atlas distinguishes this inference from experiments that directly support reversibility.
Published papers, patents and preprints all contribute to the counts but have separate labels. A preprint is a manuscript shared before formal peer review. A patent disclosure is not scientific peer review. A “published” label does not mean we have independently verified or reproduced the result. Each TF is counted once in the headline breakdown: published evidence takes precedence over patents, then preprints. A TF labeled “published” may also have patent or preprint sources. Zero “patent only” TFs does not mean that no relevant patents exist.
AI can misread measurements, tables and experimental methods. Missing full papers, inaccessible supplementary data and incomplete searches leave gaps. No binder recorded means we have not identified qualifying evidence here; it does not prove that no binder exists. Findings may be corrected as evidence changes.
Which TFs have drugs?
For each TF with a qualifying binder, we separately look for small-molecule drugs that bind that TF and are approved or being developed in human trials. Sources include trial registers, official drug information and reports from researchers or the companies developing the drugs.
Approved means a drug has regulatory approval in the country or region stated. It need not act only on that TF. Active clinical development requires evidence of an ongoing trial or current development program. Older human trials without a verified active program are labeled separately. A planned trial is not necessarily open to participants.
Secondary target means an approved or experimental drug also binds the TF, but that TF is not the established target for its clinical use. We include small molecules that bind a TF and help the cell break it down. We do not count treatments that act only indirectly on the TF.
The clinical drug may differ from the compound supporting the binding count; its clinical status does not establish its binding strength. Drug examples are not a complete inventory. “No clinical program identified” describes our search result, not proof that none exists.
What qualifies as a disease link?
We include only two kinds of evidence. Established genetic disease relationship means ClinGen or a GenCC contributor has rated the evidence Strong or Definitive: substantial evidence that changes in the gene can cause a particular disease. Definitive additionally requires that the relationship has stood up over time. These are source classifications, not new assessments by the atlas.
CGC cancer driver means Cancer Gene Census reports a role for alterations in the gene in cancer. We include all CGC tiers and do not distinguish between them. Records come through Open Targets, with its release identified in each TF’s source details. This is not an independent download of the latest COSMIC release. Disease labels reflect the source’s reported relationships; a cancer-driver label does not establish that targeting the TF will treat that cancer.
We exclude studies that link common genetic differences statistically to disease (GWAS), computer-predicted cancer drivers, changes in gene activity alone, laboratory experiments alone, and literature associations without a qualifying assessment. Moderate, Limited and Supportive assessments do not qualify. Clinical features and traits are also excluded. Disagreements between source assessments are flagged where identified.
Disease filters apply to the map, list and timeline. Each TF is counted once. Related diagnoses may have different names; some disease groups remain unclassified. “No qualifying disease record” means no relationship meeting these rules is recorded here, not that the TF has no role in disease. A disease link and a potent binder do not show that targeting the TF will be safe or effective.
Sources and acknowledgments
Evidence comes from scientific papers, preprints, patent disclosures, drug approvals and human trial records. Each TF entry links to the evidence used. The section below describes how these sources are searched.
We also use the IUPHAR/BPS Guide to Pharmacology and the studies it cites. Unit conversions and uncertain details are noted with individual results. Guide to Pharmacology content and database licenses: CC BY-SA 4.0 and ODbL.