The Biotech Analyst Speaks
Sunya Bhutta: Can you perform this analysis with other disease classes—such as CNS disease, autoimmune disease, cardiovascular disease or is this analysis restricted only to cancer?
Dr. Razvi: This analysis is scalable, i.e., it can be performed with virtually any disease class for which exists a body of scientific literature. The scientific literature en bloc is downloaded and subsequently analyzed using the computational tools we have developed.
Jose Carlos Diez-Masa: A very interesting study. Most of the molecular entities associated to cancer provided in the study are genes. Could you provide information about molecular entities related to cancer that are proteins?
Dr. Razvi: Yes, most of the elements we have provided are not only genes but also proteins—examples being PSA for prostate cancer, CA125 for ovarian cancer—which are both protein biomarkers.
Patrice Bartell: How does one test your predictions—your predictions of biomarker signatures associated with specific cancers?
Dr. Razvi: This is an excellent question. Indeed our predictions based on the en masse analysis of the scientific literature lend themselves to be experimentally validated based on both retrospective as well as prospective analyses using patient samples as well as clinical data [associated clinical records]. In this manner, the predicted biomarker signature[s] can be tested with patient samples.
Mohan Karkada: Dr. Razvi, very interesting and educating report. I guess the bias is reduced a lot when you incorporate findings from so may published reports. Is there a full report, for example, similar to the example of breast cancer shown in the report, how does ovarian cancer biomarkers look like? Just curious!
Dr. Razvi: Yes, we’ve been performing the analysis with lots of different disease classes. We present the analysis with breast cancer in this report. We have generated a full report wherein we have performed this analysis with a list of disease classes and predicted biomarker patterns.
Melinda Kucsera: How does the analysis of the entire body of scientific literature in the cancer biomarkers field allow you to make predictions regarding the association of a particular set of biomarkers with a particular cancer class or classes?
Dr. Razvi: This is an important question. Essentially, the analysis we have performed allows us to view and interrogate at once every single publication in the cancer biomarkers space. Therefore we are not looking at one anecdotal report which may be an anomaly or may have experimental bias. By analyzing the entire database of publications we can identify those trends and "markers", which are found in the many settings thereby reducing the variance/bias associated with looking at only a few publications. By viewing the space in this manner, we can identify patterns, which allows us to predict potential patterns of biomarkers associated with a particular biology/disease class.
Tamlyn Oliver: Why do you believe that PubMed is a good source of publications for this analysis? Are there any other more relevant databases?
Dr. Razvi: PubMed is the de facto standard for publications in the life sciences space. Therefore, we believe that this starting material provides the most content of relevance for us to analyze/interrogate. Also, we have performed our analyses with patent databases, but those reflect the patenting landscape and not the research/publications space.