Precision oncology emerged when genomics and new therapies converged. Genomics grew exponentially, fueled by next-generation sequencing, and gave insights into cancer physiology. This knowledge enabled the rational design of targeted therapies. Precision oncology draws from an understanding of cancer biology and an increasing collection of biomarkers and associated therapies to choose the best treatment for each patient.
Despite the undeniable success of precision oncology in multiple cancers, difficulties remain. Matching a cancer mutation to a single drug has limited utility. Many drugs that succeeded in phase I/II trials failed later phase III trials. Even for effective therapies, predicting who will benefit is difficult. Some have voiced concerns that precision oncology is not feasible outside a few well-publicized examples. Here we suggest a few ways to go forward.
In the first publication, we showed that it is not enough to consider pathogenic mutations of the EGFR gene to predict response to EGFR-directed treatment. Many patients present with concurrent mutations in other genes. Most of these mutations have no apparent effect on treatment response, but some can cause primary resistance via cellular pathways which bypass the drug's action. The second publication showed that lung adenocarcinoma with a low allelic frequency of the EGFR mutation is associated with shorter progression-free survival. The low allelic frequency is an indirect sign of subclonal expansion, meaning that the tumor is heterogeneous. Cancer cells which do not carry the mutation are less likely to respond to therapy.
The third work exemplified the use of copy number analysis in addition to DNA mutation testing. Based on an exceptional case, we presented a hypothesis to explain benefit from treatment with palbociclib and suggested how this can be applied in future trials. In the fourth publication, we developed a multi-cancer biomarker of liver metastasis from gene expression data. Pan-cancer biomarkers are difficult to identify but can reveal underlying biology. Gene expression biomarkers are particularly sensitive to batch effects and other sources of bias and are currently under-utilized.
Several challenges remain. As we showed in the first two publications, the cancer co-mutation patterns and tumor heterogeneity are challenges for the implementation of precision oncology. We believe that the accumulation of data, the integration of algorithmic decision support, and the implementation of new assays will help overcome these difficulties. In accordance with this view, our future projects are focused on biomarker discovery and tumor classification, most recently with the integration of single-cell RNA sequencing and digital pathology.
