Signal types.
Whel organizes evidence into four research arms. Select a signal type to read how it works, what its sources are, and what to look for.
Direct Research is the most literal arm: it pulls studies and trials that were explicitly designed to investigate the condition in question. If a researcher set out to study endometriosis, or registered a clinical trial for PMDD, it belongs here. This arm is intentionally sparse for most conditions in the database, and the sparseness is itself data. Endometriosis affects up to 10% of women of reproductive age, yet the direct research arm returns a relatively small number of results, most of them recent. That is not a gap in Whel. That is a reflection of how little targeted research exists.
Whel pulls Direct Research signals from two sources. PubMed (via the NCBI Entrez API) is the primary database for published biomedical literature, maintained by the National Library of Medicine. Searches are condition specific and filtered for relevance to drug or therapeutic intervention. ClinicalTrials.gov (via the ClinicalTrials.gov REST API v2) is the NIH registry of publicly and privately funded clinical studies. Whel captures active, completed, and recruiting trials, including trial phase, intervention type, and enrollment status. Each signal is analyzed by a language model for evidence strength classification: Strong, Moderate, Emerging, or Exploratory.
A concrete example: several small randomized controlled trials have examined melatonin supplementation for endometriosis related pain. These appear in the Direct Research arm because they were specifically designed to investigate endometriosis. The evidence is preliminary, with small sample sizes and limited follow-up, but the signal is there, and it is the kind of signal that should be informing the next generation of trial designs. The fact that melatonin is rarely mentioned in clinical conversations about endometriosis treatment is exactly what this arm is designed to make visible.
What to look for: A small number of strong evidence signals for a given condition suggests it is understudied at the trial level. A cluster of preliminary signals may indicate an emerging research area. The evidence strength label on each card reflects study design and sample size, not just whether results were positive.
Inclusion criteria
The highest-confidence category carries the highest bar. Minimum requirements: at least one peer reviewed human study with clearly identified population, drug, outcome, and effect direction. Signals are excluded if they are mechanistic only with no human data. Preferred: at least one prospective study, trial, or meta-analysis. Quality criteria prioritize replication and outcome relevance over citation count; a highly cited older paper with no replication is not equivalent to two recent independent studies with similar findings.