The Value of Molecular Disease Models for Cancer Treatment Planning
Smruti Vidwans, PhD, Chief Science Officer at CollabRx
Q: You have been notified that your article entitled “A Melanoma Molecular Disease Model” is among the top 10% of most cited articles in 10 years at PLOS ONE. Congratulations. What is the essence of that paper, why it is important, and what has changed in the past 6 years?
A: Molecular testing of cancer, especially late stage, is increasing in a clinical setting. For it to be relevant, molecular profiles have to be linked with potential treatment options (approved or investigational drugs) and clinical trials. In the paper alluded to above, we 1) defined “actionable” molecular subtypes of melanoma and 2) linked them to clinical trials and drug/ drug classes on the basis of published literature.
Subtypes were defined at the level of pathways and biomarkers/genes rather than individual variants. For example, subtype 1 was defined as having an aberrant MAPK pathway, either by itself or in combination with the AKT/PI3K or CDK pathways. Subtype 1.1 provided more granularity and was characterized as having a mutation in a specific component of the MAPK pathway – BRAF.
The field of precision oncology has evolved much since this paper was published 6 years ago. For example, with next generation sequencing (NGS), non-canonical and uncommon variants can now be identified. Before NGS, patients may have learned only that their cancer did *not* have the common, activating BRAF mutation – V600E. With NGS, they may now learn that their cancer, instead, has another activating BRAF variant – R462I. A classification scheme today must support a more nuanced molecular profile.
Over the last few years, there has been an explosion of clinical and preclinical data around the therapeutic impact of molecular biomarkers and specific variants. For the BRAF R462I variant, there may, in fact, be studies that suggest therapeutic relevance. But if not, is this variant inactionable? Not necessarily. There might be learning points from “related” variants that could be applicable. “Related” variants could include those that 1) have the same functional effect on a biomarker (activation of BRAF), 2) reside in the same protein domain (BRAF kinase domain) or 3) lie in the part of the gene shown to have functional impact (EGFR exon 19).
Therapeutically relevant data may be in a cancer different than the patient’s diagnosis. Should data from another cancer be used in making treatment decisions? If so, under what circumstances? In another publication (Genomically Driven Tumors and Actionability across Histologies: BRAF-Mutant Cancers as a Paradigm), we used BRAF as an example to posit that data from one cancer could inform choice of therapies in another cancer, perhaps as part of combinations.
If predictive data is available for a specific drug (say vemurafenib for BRAF V600E), should that be the only drug under consideration? What if the patient has already been treated with that drug and is resistant or that drug is not available to that patient? Could “similar” drugs (say, other BRAF inhibitors) be used? If so, what is the heuristic?
Today NGS provides physicians with a wealth of molecular data that could inform treatment for their patients’ cancer. However, in many cases, this may depend on leveraging data about “related” variants, data from other cancers, and data about “related” drugs. Risk tolerance of individual physicians toward using these types of data may vary. User-friendly reporting tools that map the variant/cancer/drug relationships leveraged in formulating potential treatment options may help physicians quickly evaluate whether to use a particular variant to inform the choice of treatments.
A: Molecular testing of cancer, especially late stage, is increasing in a clinical setting. For it to be relevant, molecular profiles have to be linked with potential treatment options (approved or investigational drugs) and clinical trials. In the paper alluded to above, we 1) defined “actionable” molecular subtypes of melanoma and 2) linked them to clinical trials and drug/ drug classes on the basis of published literature.
Subtypes were defined at the level of pathways and biomarkers/genes rather than individual variants. For example, subtype 1 was defined as having an aberrant MAPK pathway, either by itself or in combination with the AKT/PI3K or CDK pathways. Subtype 1.1 provided more granularity and was characterized as having a mutation in a specific component of the MAPK pathway – BRAF.
The field of precision oncology has evolved much since this paper was published 6 years ago. For example, with next generation sequencing (NGS), non-canonical and uncommon variants can now be identified. Before NGS, patients may have learned only that their cancer did *not* have the common, activating BRAF mutation – V600E. With NGS, they may now learn that their cancer, instead, has another activating BRAF variant – R462I. A classification scheme today must support a more nuanced molecular profile.
Over the last few years, there has been an explosion of clinical and preclinical data around the therapeutic impact of molecular biomarkers and specific variants. For the BRAF R462I variant, there may, in fact, be studies that suggest therapeutic relevance. But if not, is this variant inactionable? Not necessarily. There might be learning points from “related” variants that could be applicable. “Related” variants could include those that 1) have the same functional effect on a biomarker (activation of BRAF), 2) reside in the same protein domain (BRAF kinase domain) or 3) lie in the part of the gene shown to have functional impact (EGFR exon 19).
Therapeutically relevant data may be in a cancer different than the patient’s diagnosis. Should data from another cancer be used in making treatment decisions? If so, under what circumstances? In another publication (Genomically Driven Tumors and Actionability across Histologies: BRAF-Mutant Cancers as a Paradigm), we used BRAF as an example to posit that data from one cancer could inform choice of therapies in another cancer, perhaps as part of combinations.
If predictive data is available for a specific drug (say vemurafenib for BRAF V600E), should that be the only drug under consideration? What if the patient has already been treated with that drug and is resistant or that drug is not available to that patient? Could “similar” drugs (say, other BRAF inhibitors) be used? If so, what is the heuristic?
Today NGS provides physicians with a wealth of molecular data that could inform treatment for their patients’ cancer. However, in many cases, this may depend on leveraging data about “related” variants, data from other cancers, and data about “related” drugs. Risk tolerance of individual physicians toward using these types of data may vary. User-friendly reporting tools that map the variant/cancer/drug relationships leveraged in formulating potential treatment options may help physicians quickly evaluate whether to use a particular variant to inform the choice of treatments.
Copyright: This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
