All Articles


Multiomics has recently gained traction as a promising avenue towards making precision medicine technologies work. Multiomics is an approach that combines analyses from the genome, proteome, transcriptome, epigenome, metabolome, and microbiome to generate a personalized molecular signature to be theoretically used to inform patient treatment. I wanted to know to what extent these technologies are working, and who this approach actually helps. To illustrate, we’ll take Q bio and Freenome as two startups doing work in this space.

Both of these companies tout advanced machine learning and artificial intelligence powered insights that is enabled via the collection of a LOT of data. The idea is that earlier detection of disease will allow physicians to act sooner and save people their time, health, and money. To collect this data, Q will conduct a 90 minute exam, in which saliva, blood, and urine will be collected, along with a whole body MRI, vitals, and insights from a person’s medical record. Freenome is a blood test that predicts presence of cancer based upon RNA, DNA, and other blood signatures.

For each, the technology provides the greatest utility for those without preexisting conditions, healthy individuals with no sign of disease. Why is this? Consider someone who is obese with high cholesterol. Physicians basically know what is wrong and what interventions to prescribe. This is entirely at a clinician’s judgement but in most cases the patient will be prescribed a statin and told to eat better and be more active. There isn’t anything fancy that needs to be done here, and there was no need to do the intense screening that Q or Freenome offers.

Recent data from the NCI shows that early detection accounted for less than 5% of the overall gains in mortality. Early detection seems like a no brainer, but there are some serious hurdles that need to be met. A test would need high specificity in order to prevent false positives. Embedded in this question is also what to do about benign tumors. A highly sensitive test may pick up many small nodules that are present even in perfectly healthy individuals. In order to truly provide aggregate value, a test needs to demonstrate that it improves overall survival, not just survival from the tested indication. You might think that a test that improves mortality from a specific cancer would improve overall mortality, but this is not the case. (See the NELSON lung cancer screening trial)

Another big concern is cost. Blood based tests can be expensive to develop, and almost all startups tackling this problem are propped up by enormous VC rounds. Q bio offers their test for $3500, which isn’t covered by any health insurance. My concern with these tests is that they are designed for people that already care deeply about their own healthcare, are already mostly healthy, and who have the money to pay out of pocket for high quality care. It falls into the same hole as other precision medicine technologies. The subset of the general population that this technology benefits is small, and will stay small for some time.

To conclude, I don’t have any advanced degrees, but I’m unconvinced by current literature. Theres significant bioplausibility here, but it needs to backed up by robust clinical data. The most promising data has come out from Thrive Earlier Detection and the intervention led to a surgical intervention in 9 out of 10,006 women, and remember this doesn’t have a control arm. Projecting when these technologies will bear fruit is a tricky task. Yet, these types of startups have received considerable funding. They’re very trendy. Still, it makes sense as a moonshot, let’s all thank VCs for taking on the risk.

Published May 5, 2020

Harvard-MIT PhD Student