AI medical trial: a natural-history study collecting medical data on genetic conditions to build "advanced data analytics" (AI) tools (NHGRI observational study) — a clinical trial (ClinicalTrials.gov)

1Why genetic conditions are hard to diagnose

This study (NCT05657405) is a natural-history study that gathers medical and genetic information from people with genetic conditions — diseases that the genes a person is born with can cause — and their relatives. Its goal is to help develop "advanced data analytics" tools for better analyzing and understanding genetic data.

2The problem the study starts from

Per the registry summary, the motivation is this: genetic diseases are rare, but they can have a big impact on those they affect. Researchers have already made great strides in understanding how some genes cause disease, but they want even better tools to analyze and understand genetic data — and building such tools requires gathering health and genetic data from many people.

This study is best understood as that data-foundation effort. Participants come to the clinic for up to four days (some may participate via telehealth), and the tests vary with each person's health issue: genetic testing from blood and saliva, imaging scans such as X-rays, heart tests, and photographs or video/audio recordings of facial features, skin changes, or other effects of the condition.

3Finding the causal variant in a vast genome

In genetic conditions, the causative change is often buried in a vast amount of genomic information, making it hard to tell which variant actually causes disease. This is where AI and other advanced data analytics are expected to help — by finding informative patterns across large genetic and clinical datasets and offering clues for hard-to-diagnose cases.

Note, however, that the study does not assert the performance of any new analytic tool; it is an observational study that first carefully assembles the data on which such tools can be built.

4Gathering the data before building the tool

The purpose here is not to measure the performance of a finished analytic tool. It is to prepare the ground on which one can be built.

1. 1Collect from patients and familiesUp to four days of visits covering genetic testing, imaging, cardiac studies, photographs and recordings
2. 2Build the data baseAccumulate health and genetic data from many people
3. 3Grow the analytic toolsBuild what can tell which change among vast genomic data causes disease
4. 4Reach a diagnosisObtain leads in cases that resist diagnosis

Genetic disorders are individually rare while heavily affecting those who have them, and the causative change is often buried in vast genomic data. What distinguishes this study is that its main purpose is collecting data to grow the analytic tools themselves rather than testing a particular drug or test. It does not assert the performance of any new tool.