Do you think we'll ever get to a state in which perfect algorithms will allow us to deduce a "virtual clinical outcome" on the basis of genotypic and phenotypic data?
Moderator: The next question is from Jonathan. And Brian and maybe David Katzenstein, you can both comment on this. Do you ever think we'll get to a state in which we can get almost a virtual clinical outcome, that is, can we ever get to an algorithm for interpreting genotypic and phenotypic data?
Brian Conway: I think it's certainly something that we need to go towards, and the ideal report for me of a resistance test result would include some information on what would be a preferred therapy as long as the person who is reading the genotypic and phenotypic resistant result is also provided
with a drug history of the patient.
And the analogy I would make is with a CT scan report of the head or the chest or the abdomen or whatever. There includes a description of what was found, and it includes a best interpretation of what this might represent, given the clinical information that the radiologist is provided. So in an ideal situation, the person who is interpreting the result should be given the genotypic data, the phenotypic data in terms of fold resistance, the drug history, and provide an interpretation the same way as an opinion is provided at the bottom of a CT scan result. Now, whether this will always be perfect is clearly not possible, but it certainly will help point people in a direction where they feel more comfortable making a decision themselves or saying well, this is too complicated, I need to refer to the one specialist in my area that will help me make the right decision.
David Katzenstein: As an expert virologist who loves to do that, of course I think that's the best thing to do. But in fact, I think informatics is rapidly going to overtake us, not that the current virtual genotype is nearly approaching the degree of sophistication we want, but at least I think the vision that with the accumulation of sufficient data, we will actually be able to ask the question, how does a patient do clinically with a given phenotype or genotype given a given set of drugs. That is really just an information question. Can we actually accumulate between clinical trials and clinical practice a sufficiently uniform amount of data to understand how it is that in fact the real biologic response we're interested in, which is a decrease in virus load, can in fact be estimated from past experience. Bob and I are certainly working with very small data sets to ask that question, and many of us have tried to analyze studies doing exactly that. So I think a virtual clinical outcome database will be possible. It certainly now is not.
