3billion has recently announced 3Cnet, a pathogenicity prediction tool. Its performance surpasses existing tools.
What is the significance of that, and what could it mean for the diagnosis of rare genetic diseases, moving forward?
Let’s take a quick look and break it down into more simple terms.
- 3Cnet is a pathogenicity prediction tool, based on recurrent neural networks (RNN), which can detect disease-causing variants in humans
- It has been in development for roughly a year and a half
- Achieved a top-1 recall of 14.5%, surpassing REVEL’s leading top-1 recall of 6.5%
- This means that 3Cnet can predict disease-causing variants of genetic disease patients with a sensitivity ~2.2 times greater than the current-leading tool
- ‘3C’ is derived from the "three C’s" approach utilized; each
Crelates to how the model was trained
- Conservation - derived from utilization of UniRef30 for simulated mutation data
- Clinical - derived from utilization of ClinVar database
- Common - derived from utilization of gnomAD database
What is ‘pathogenicity prediction’?
Much like its name suggests, pathogenicity prediction is the prediction of pathogens, or organisms which can cause a disease. This is where we get the term
disease-causing variants from.
By applying various techniques, we are able to make accurate predictions, and streamline the diagnostic process. 3Cnet is able to look at a large number of variants, compare them against other variants, and determine which ones are malicious (disease-causing).
How does that benefit patients?
- 3Cnet has potential to increase diagnostic rates, by streamlining the process of diagnosis using AI to a high degree
- The faster and more efficient the diagnostic process is, along with high accuracy, the faster a patient can be diagnosed
- The more data it looks at, the more accurate its predictions become, topping out at near-100% accuracy
- 3Cnet looks at a long sequence of data, and by applying AI techniques, picks out specific variants which it predicts are disease-causing
What is the future of 3Cnet?
- Currently, 3Cnet can look at a
window, or sequence, of 201 amino acids around the variant site; this will be increased so that a larger sequence is able to be handled
- By increasing sequence size, 3Cnet widens its scope, providing more coverage for the patient
- Improvements to speed and performance are underway, to produce results faster
3billion Cares About YOU
WES Is Our Specialty
- We specialize in WES, and often execute trio-based WES testing for patients
- With WES, we target 1% of the human genome, the protein-coding region, which contains 85% of disease-causing mutations
- All testing we do is in accordance with ACMG/AMP guidelines
Free Lifetime Reanalysis
- Unlike alternative options, 3billion does not charge for reanalysis… ever
- Our AI-driven reanalysis runs 24/7