CHICAGO (GenomeWeb) – Informaticists from Weill Cornell Medicine in New York — with assistance from Memorial Sloan-Kettering Cancer Center — have compiled, catalogued, and standardized information into a digital pharmacogenomics variant repository as they look to prevent cardiotoxicity from chemotherapy treatments.
The repository, called the Cardiotoxicity of Chemotherapy Knowledgebase, or CATCH-KB, links to existing, open-access resources, including the Pharmacogenomics Knowledgebase (PharmGKB), the Kyoto Encyclopedia of Genes and Genomes (KEGG), and the Genome-Wide Association Study Catalog in hopes of spurring new hypotheses for preventing chemotherapy-induced cardiotoxicity.
"We're hoping to incorporate our information that we've gathered in CATCH-KB into [Weill Cornell's own] Precision Medicine Knowledgebase so it can inform those results as well," explained Maggie Morash, a research aide in the teaching hospital's Division of Health Informatics. "We're hoping to be able to do some association studies with that cohort to see if any of the variants that have been identified in CATCH-KB show an association with chemotherapy-induced cardiotoxicity in that population, since a lot of the variants that have been identified really do need further validation." Morash said.
Morash presented a paper on CATCH-KB last month at the American Medical Informatics Association's annual Informatics Summit in San Francisco.
Chemotherapy-induced cardiotoxicity is a rather narrow field, but it is something that Morash's mentor Jyotishman Pathak, chief of health informatics at Weill Cornell, had been interested in from his previous job at Mayo Clinic.
"It is very specific," Morash said, "but it is something that needs specific attention." The integration with other, broader resources fills in some important gaps for researchers, she explained.
"There are a lot of endpoints that are really important for cardiotoxicity studies that aren't necessarily as important for other pharmacogenomics studies," Morash noted. For example, it is important to widen the follow-up period when researching cardiotoxicity because the condition can present long after the cessation of treatment.
"The other thing that we wanted to do was to give some pretty detailed population data, which PharmGKB does as well to some degree," Morash continued. "Hopefully that level of detail can identify patterns in populations, because there have been mixed results if it's specific to different populations."
In building CATCH-KB, Morash and colleagues searched PubMed for literature that has studied relationships between germline genetic polymorphisms and chemotherapy-induced cardiotoxicity. At the time they submitted their paper to AMIA in January, they had identified 49 studies on humans and 50 more on nonhuman model organisms.
Since acceptance of the AMIA paper, the database has grown to about 60 human studies.
Her team receives alerts from PubMed whenever there is a relevant new paper published, including for those that are non-germline or tests on model organisms.
"The focus is still very much on the human studies, and we feel pretty confident that it's pretty extensive in that regard," Morash said. Eventually, she wants to move into model organisms, with the goal of making CATCH-KB more useful for drug discovery, but there simply was not enough time before the AMIA submission deadline to analyze the studies on other organisms.
"In terms of trying to help discovery, we've been trying to incorporate more model organism papers so we can identify genes and variants that haven't been evaluated at all in humans," Morash said. "If there's a gene that someone has looked at in a model organism that seems like it's associated with cardiotoxicity, and it's known to be in a pathway that mechanistically could be important to it, then we want to provide that [information] to help discovery."
Part of the reason for the deliberate pace of expanding CATCH-KB is that the team has to manually map search results to standardized ontologies, including SNOMED-CT and RxNorm. "We really think it's important for it to be standardized so it can be readily integrated with other resources," Morash said.
Morash said that she and her colleagues are working with application programming interfaces from the National Center for Biomedical Ontology's BioPortal, but they first must manually extract the terminology from each paper, then feed that information into the API.
The team is considering several techniques to automate the intake of new material to "identify elements of interest in them using some data-mining techniques and also to link more fully with other resources," Morash explained. CATCH-KB currently provides just a link to original sources in places like PharmGKB, KEGG, and the GWAS Catalog. "We want to be able to automatically extract the relevant information to more readily include it all in one place for the user," Morash said.
Nearly all of their work so far has been within the walls of Weill Cornell. Morash admitted that she has not actually tracked traffic to the CATCH-KB website, and said that the AMIA presentation was the first real overture to the wider bioinformatics community. The forthcoming connection to the Precision Medicine Knowledgebase should help with dissemination.
For now, CATCH-KB is designed for research only. Morash said that there simply is not enough evidence on cardiotoxicity from chemotherapy yet for such research to inform clinical care.
"The hope is that … we can start to include this in the Precision Medicine Knowledgebase and that the information that we have could be included in annotations that they provide in their summary reports when they return their sequencing results for patients," Morash said.
At the moment, Morash does not believe the annotations could be listed in sequencing reports with a high degree of confidence. However, she suggested that CATCH-KB could provide annotations for reference in case a specific patient exhibits other cardiotoxicity risks.
"The first step is … to facilitate further studies to validate the variants, and then [later] incorporate it into clinical workflows so it can be helpful at decision points," she said.