Hot questions in this board:
Hot questions in other boards:
Well, I just heard Craig Ventor describe what he was doing and it sounded like a good start: using supercomputers (AI) to find correlations between people/patient's biomarkers and their phenotypes or diagnoses. Until the data set is sufficiently large, I don't believe that clinical genomics will move past the relatively few biomarkers known in oncology.
I agree can with you on that. I do believe though that the first task for novel discovering using AI is that of creating learning sets using even more advanced technological tools such as NLP which provides context based data extraction. Using such curated data for training, we train a model that allows for creating result-sets immediately and on the fly for assembling a rich corpus and generating semantic layers of information. This approach has the potential to overcome the initial cited problem of understanding clinical genomics or even cancer biology so to speak.
So far havn't designed our supercomputers to take in all the biomarkers data and havn't given them the power to answer and corelate what we want from them. hence our first step is to throw open the back doors of google's stored biomarker discoveries , and then to train AI on retrieving the clinical apllications of proposed biomarkers !