06 Jan canSAR Database Analysis Speeds Identification of New Cancer Targets
MedicalResearch.com Interview with:
Dr Bissan Al-Lazikani
Team leader in computational biology
The Institute of Cancer Research
Medical Research: What is the background for the canSAR database? What are the main uses for the tool?
Dr. Al-Lazikani: Drug discovery is a difficult, time consuming and expensive venture that frequently ends in late stage drug failures – especially in oncology.
As with any complex venture, decisions throughout the drug discovery pipeline can be empowered by having access to the right information at the right time. But for drug discovery this means bringing together billions of experimental data from very diverse areas of science spanning genomics, proteomics, chemistry and more.
We developed canSAR to help guide our own drug discovery efforts by integrating these huge, diverse data and by analysing the data and deriving hidden links and knowledge from them. This means that we can answer questions in minutes that would have taken weeks using previously available public resources. But, more importantly, canSAR analyses and links these data in a way that allows us to derive knowledge that was hidden before. For example, one of the main ways canSAR is used is to help select the best druggable targets for drug discovery. Using canSAR we were able to uncover many druggable cancer proteins that were previously overlooked, and we are delighted to see that several of these proteins are now the subjects of drug discovery and development projects both by us and by others.
We took the decision to make canSAR publicly and freely available because we believe that cancer drug discovery is a vast challenge that requires openness and data sharing worldwide. It has been embraced by the community is being used by tens of thousands of cancer scientists worldwide, both in academia and industry, to generate hypotheses for experiments and select targets for drug discovery.
Medical Research: How does the new 3D version use artificial intelligence to design new cancer blocking drugs?
Dr. Al-Lazikani: The key advance in this version is that we take a holistic view on all human proteins in terms of their function and how suitable they are for drug discovery. It’s important to remember that not all cancer-causing proteins lend themselves naturally to drug discovery and development; so it is imperative that we select the correct point in the cancer cell to target before potentially investing in the wrong target.
Having developed canSAR and classified the data within it, we are able to train computers to make predictions about what would make a good cancer drug target. We trained our algorithms to assess the three-dimentional architecture of a protein; its chemical interactions; as well as its communication patterns with other proteins in the cell. The algorithms then make a prediction about how suitable a protein is for drug discovery. The fact that we are using artificial intelligence means that the algorithms can make predictions on data and targets that they have never seen before and allow us to explore completely new areas of biology and drug discovery, but with the risks reduced. This is particularly important because we need new drugs acting in different ways to overcome resistance to well-established cancer drugs. Doing this at the massive scale we can achieve in canSAR means that researchers do not need to be limited to narrow, well-established pathways or target groups.
Medical Research: What future research are you planning?
Dr. Al-Lazikani: We have done a good job streamlining the information flow from early target discovery to later stages in the drug development pipeline. But drug discovery and development are not linear. New findings in the clinic should all go back to inform all drug discovery stages. This relies on scientists coming across important data, or having the right links to the right clinicians at the right time. We believe that a two-directional flow of knowledge between preclinical research and the clinic is essential for future drug discovery, patient stratification and overcoming drug resistance. We will focus our efforts on integrating knowledge from clinical trials and developing new algorithms to make predictions on likely patient cohorts, drug responses and drug resistance.
Joseph E. Tym, Costas Mitsopoulos, Elizabeth A. Coker, Parisa Razaz, Amanda C. Schierz, Albert A. Antolin, Bissan Al-Lazikani. canSAR: an updated cancer research and drug discovery knowledgebase. Nucleic Acids Research, 2015; gkv1030 DOI: 1093/nar/gkv1030
Dr Bissan Al-Lazikani (2016). canSAR Database Analysis Speeds Identification of New Cancer Targets