Clinical Oncology

Knowledge Management System (KMS)

The GO KMS enables precision medicine by allowing users to aggregate and analyze biomarker-based data in the context of a library of trusted data sources that can also be extended with institution-specific interpretations as desired.

The KMS integrates molecular and clinical information, providing oncologists with powerful decision-support tools.



GO Solutions are built around our KMS.

  • GO has partnered with the My Cancer Genome (MCG) team at Vanderbilt to develop the KMS, while the MCG team is responsible for the content
  • The KMS “understands” where genes are on chromosomes, what exons make up a gene, how those exons make-up transcripts, which proteins are coded for by the gene and its domains, and what the role the protein plays in cellular pathways
  • GO has built a powerful biomarker model based upon this framework
  • The KMS also contains ontologies for diseases and therapies
  • Assertions (i.e. rules) added to the KMS leverage these ontologies so that rules can represent knowledge in the form used by oncologists (e.g. NCCN guidelines)
  • The licensable “curation interface” to the KMS is used by the MCG team to create assertions from guidelines and literature, backed by a full evidence model. GO has an exclusive commercial license to all MCG content
  • The KMS is both a decision support tool and a big data tool. Once licensed, its APIs can be used to analyze a single case (e.g., using the GO Tumor Board Application) or thousands of cases simultaneously (e.g., querying for “Cases-like-mine” in the Tumor Board and/or GenomAnalytics).



Knowledge is a set of manually curated assertions or “rules” that allows the operation of an inference engine that provides biomarker and disease-based recommendations of therapies, prognoses, and clinical trials. A full evidence model backs these assertions, allowing a curator to attach references, set a level-of-evidence, and even highlight the exact section of a reference that was used to create an assertion. This evidence is available to the end-users of the system.

Assertions are Boolean logical statements, built using the operators “All”, “Any” or “None”. Assertions can be multi-level (i.e., assertions may contain assertions). For example, an NCCN prognostic guideline for AML can be represented as follows:



That is, the rule only fires if the sample being analyzed contains either of the two ASXL1 mutations, a Normal Karyotype (as defined by ISCN) and none of the three mutations listed at the bottom right of the figure above. The KMS “understands” genes, proteins, protein domains pathways and biomarkers through set of inter-related ontologies. In addition, note that criteria for rules can be specified at a higher level than a Ref-Alt or HGVS alteration (e.g., “NPM1 Exon 11 Mutation” or FLT3 ITD).

Through the use of the GO Biomarker bridge, the KMS allows for curation at a high level but triggers based on the actual alterations found through NGS, IHC, cytogenetic analysis, FISH, etc. For example, assertions can refer to EGFR Exon 19 deletions, and these rules trigger against any in-frame deletion within the boundaries of the 19th exon of the EGFR gene:



This technology allows curators and oncologists to deal with hundreds of rules instead of millions of rules (i.e., if the system required a rule for every alteration ever observed), providing a manageable data set that can be both human-curated and maintained.

The GO framework supports the parsing and interpretation of NGS variants as well as biomarker concepts. GO HGVS parser can decipher all variant types, including insertions, deletions, substitutions, intronic, splice and promoter markers. The biomarker model is not limited to simple genetic alterations: fusion data, copy number changes, protein expression information acquired via IHC (ER, PR, MSI, PD-L1, etc.) can also trigger rules. The KMS also contains a full cytogenetics engine that parses out the individual aberrations from ISCN strings and uses those concepts, e.g.” i(17)(q10)” implies “del(17)(p10)” in triggering rules content. APIs exist for entering all these kinds of data into the KMS. In addition, users can input the data through the UI in any number of ways, including manually adding markers, uploading .vcf files, or Foundation Medicine XML file parsing.

This assertion-based approach is also applied to clinical trial matching. The My Cancer Genome team has curated all cancer trials listed in Clinicaltrials.gov that contain biomarker references (about 2500 trials, currently). Trials eligibility criteria are based on a combination of biomarkers and disease (and will soon support criteria based on prior interventions and other clinical criteria). For example, a multi-arm trial could contain the sets of inclusion and exclusion criteria as shown on the left of the following figure, which for a real trial (e.g., NCI Match Arm H) would look like what is shown to the right of the figure, below.



This assertion-based approach is also applied to clinical trial matching. The My Cancer Genome team has curated all cancer trials listed in Clinicaltrials.gov that contain biomarker references (about 2500 trials, currently). Trials eligibility criteria are based on a combination of biomarkers and disease (and will soon support criteria based on prior interventions and other clinical criteria). For example, a multi-arm trial could contain the sets of inclusion and exclusion criteria as shown on the left of the following figure, which for a real trial (e.g., NCI Match Arm H) would look like what is shown to the right of the figure, below.




Schedule a demo