“We have to bring “genome-drug” interactions to (physicians’) attention just as we currently bring “drug-drug” interactions to their attention.”
Adverse drug events account for over 700,000 deaths each year, and nearly 30% of these are attributed to interactions of drug combinations. Public databases curate hundreds of thousands of gene variants linked to disease risks every year. Mining these diverse sources could help us learn how genetic variations, drug targets and clinical parameters come together to influence human health. Using computational tools to utilize this wealth of scientific data effectively is something we’ve discussed on the blog earlier as well.
Beginning at the “intersection of molecular biology and medical informatics” over ten years ago, Russ Altman is the founder of PharmGKB (PharmacoGenomics Knowledge Base), a database that curates and disseminates information about gene-drug-disease relationships. The professor of bio-engineering, genetics and medicine at Stanford University is also on the Scientific Advisory Board at NextBio, and spoke to us about genomics and the future of medicine.
NB: What got you interested in bioinformatics initially?
I have always been interested in biology and computers. The word “bioinformatics” did not exist when I was hired, and so my offer letter said I would work at the “intersection of molecular biology and medical informatics”! Pharmacogenomics came much later-Dr. Kathy Giacomini from UCSF came to visit and recommended I get involved. When I realized that my interests in medicine, pharmacology, informatics, genomics all combined in this field, I was hooked.
NB: How did PharmGKB come about?
The NIH had a workshop and subsequent call for proposals to encourage the development of pharmacogenomics research in academia in 1999-2000. They wanted to fund research, but they also wanted to fund a knowledge base to catalog how human genetic variation impacts drug response phenotypes. I thought we were a good match, and so we wrote a proposal for PharmGKB, and were successful. We have now existed for 11 years.
NB: Are there any particular success stories that have resulted from PharmGKB that stand out for you?
The IWPC (International Warfarin Pharmacogenetics Consortium) is one of our biggest success stories, because we helped bring 21 groups together that are using genetics to answer a simple question about how to dose warfarin. We have set up similar consortia in other areas, and they are working hard to finalize their results and report them. Another big success has been our work doing the first clinical annotation of a complete human genome (published in Lancet last year). Finally (and most obviously) the PharmGKB website is getting 30-40,000 unique IP hits every month, and all those users are hopefully making small advances in their science by using our resource.
As whole genome sequencing becomes less expensive, I think we can contemplate measuring the whole genome of patients in clinical trials (particularly drug trials) and use the information in the genome as a standard set of “covariates” to do the statistical analyses of response. I think that systems pharmacology will also allow us to have a more detailed understanding of drug response, which should lead to better triage of lead compounds.
NB: What are the challenges with effectively translating genomic advances into clinical practice?
There are two different challenges. For pharmacogenomics, I think it is not easy but more straightforward: we have to implement warnings to physicians in their order-entry systems to bring “genome-drug” interactions to their attention just as we currently bring “drug-drug” interactions to their attention. The physicians will then be able to assess the advice and decide whether to use another drug, adjust dose or exercise increased vigilance. For disease risk, it is much trickier because we don’t really know what the long term risks are for patients with risk variants, and we are not able to explain most of the risk. Also, physicians will need to have a much deeper understanding of the interpretation of disease risks, and I think it is harder than drug advice.
NB: Do you think making data publicly available, including electronic medical records (EMR)s, could change the pace of drug discovery?
Yes, I think if we can make sure that individual patient privacy is protected (either through technical means or social means, like large penalties for violations), then I think that modern data mining and machine learning technologies have a great chance of finding very valuable knowledge in these databases, particularly if they can be linked together and integrated so that random sources of noise kind of cancel each other out. (For example, linking genetic information with EMRs from patients.)
NB: How did you get involved with NextBio?
I am collaborating with Mostafa Ronaghi, one of the NextBio founders. He showed me a demo of some of the early NextBio tools, and it looked pretty exciting, so I said “Sign me up!”
NB: Are there any specific aspects of NextBio that appeal to you as a researcher?
The focus of NextBio on biological workflows is really key. They work very hard to allow a busy researcher to quickly scan for interesting trends/hypotheses. Instead of providing overly complex power tools, they have worked hard to try to anticipate the most common user queries or actions, and provide those in a straightforward manner.SHARE