140P - Similarity-based automated evidence ranking for clinical interpretation of multigene diagnostic panels

Date 10 October 2016
Event ESMO 2016 Congress
Session Poster display
Topics Translational Research
Basic Principles in the Management and Treatment (of cancer)
Presenter Istvan Petak
Citation Annals of Oncology (2016) 27 (6): 15-42. 10.1093/annonc/mdw363
Authors I. Petak1, C. Hegedus2, Z. Binder2, M. Peeters3, C.D. Rolfo4, G. Keri5, R. Schwab2, L. Urban6
  • 1R&d, Oncompass Medicine Hungary, 1024 - Budapest/HU
  • 2R&d, Oncompass Medicine Hungary, Budapest/HU
  • 3Department Of Oncology, University Hospital Antwerp, 2650 - Edegem/BE
  • 4Phase I - Early Clinical Trials Unit & Center For Oncological Research Of Antwerp (core), Antwerp University Hospital, 2650 - Edegem/BE
  • 5Medical Chemistry, Semmelweis University, Budapest/HU
  • 6Pulmonology, Matrahaza University and Teaching Hospital, 3233 - Matrahaza/HU



Precision medicine incorporates individual molecular genetic profiles in cancer therapeutic decisions. Establishing clinical relevance of tumour biomarkers can be limited by distinct biomarker functions in different histology types, simultaneous presence of multiple biomarkers in different combinations, long-tail distribution of driver genetic alterations and the resulting limited number of similar cancer cases. Difficulty in aggregating definitive phase 3 trial data further hampers optimal delivery of tumour genomic information to the clinic especially in the case of rare mutations. These might result in conflicting evidence regarding biomarker roles and requires decision support algorithms to help cancer treatment decisions.


Here we describe a novel decision support system which is capable of dynamically aggregating and ranking scientific and clinical evidence to aid cancer therapeutic decisions.


The algorithm aggregates scientific evidence and clinical experience for an automated, adaptive ranking of anti-cancer therapies that best match with the molecular and clinical profile of the individual patient and are supported by the most relevant evidences. Input is generated by cancer molecular profiles and clinical parameters. Outcome is represented by identification of clinically relevant tumour genomic alterations and matching drugs. Compounds are ranked based on the number and merit of scientific evidence that supports the functional relevance of identified genetic alterations (biomarker or driver evidence) and their association with drug targets (target evidence) or compounds (drug evidence). Direct association between drugs and tumour histologies are also counted. Evidence is also weighted based on similarity to the given cancer case. The algorithm is used to combine and prioritize evidences based on their relevance (clinical over preclinical, direct over indirect, registered over non-registered indications) and level. Evidences are constantly updated and accumulated.


Incorporation of similarity based evidence ranking in classical evidence-based medicine enhances the delivery of genomically informed precision medicine.

Clinical trial identification

Legal entity responsible for the study

Oncompass Medicine Hungary


Oncompass Medicine Hungary


All authors have declared no conflicts of interest.