Mon Nov 04 2019
Short Term Strategy
In the on-going COVID-19 pandemic an SOS treatment is urgently needed. In the light of the current situation in the USA, where social distancing is upheld no more, efforts need to be made to protect those at risk now.
Instead of long-term vaccine or drug development, biotx.ai identifies genetic signatures indicative of the severity of a given person's symptoms with the goal of helping protect those that will be severely affected by the disease, while enabling those who will exhibit no symptoms to return to a normal life.
Due to the lack of sufficient phenotype data biotx.ai focuses its AI approach on investigating interactions of novel genes with reported features such as HLA haplotypes, ABO blood groups, and variants in the genes HBB, CFTR, CCR5, as well as APOL1 that influence infectious disease susceptibility in general.
The generated models are being validated with data from 500,000 subjects in the UK Biobank.
Long Term Strategy
COVID-19 has had a strong influence on all of our lives - from just a few cases and media playing it down in January, to hundreds of thousands of infected people in all parts of the world, lockdowns and news reporting about it 24/7.
Drug repurposing reflects a short-cut to a treatment. There are nascent efforts at using AI to repurpose existing drugs (or candidates) for COVID-19 (see https://singularityhub.com/2020/03/31/can-existing-drugs-fight-covid-19-ai-is-on-the-case/ for a summary of the idea and https://www.cebm.net/covid-19/registered-trials-and-analysis/ for an overview over the current projects).
Machine learning can predict the efficacy of an existing compound against COVID-19, based on that compound’s effect on known illnesses. But the above article questions whether the suggested compound truly works on COVID-19, as there is no gold-standard data, and concludes: ‘No one knows’.
We beg to differ. During his PhD, our CSO, Dr. Marco Schmidt, developed potential drug compounds against the the main protease (disease agent) of SARS-CoV-1. He received the prestigious Klaus-Grohe-Preis award for his outstanding work. Not only that, but in his post-doc at Cambridge, Marco’s work for the Bill & Melinda Gates Foundation focused on new anti-infectives. Further, our AI has proven to detect complex, previously unidentified genetic patterns that cause diseases - providing leads for new drug targets. The synergy between an AI that generates real new ideas and an expert that keeps the AI in check sets our approach apart:
Our AI detects complex, never-before analyzed patterns in existing GWAS and gene expression data. The results that the AI proposes will be expertly judged by Marco and his network. The AI, will, in turn, use the experts’ judgements for future predictions. This sets off a feedback loop in which the AI learns to think like an expert in the subject, but at scale. This idea of AI being checked by humans has proven to be very successful in other contexts - think Paypal’s fraud detection or Palantir for famous examples.