Identifying driver cancer genes holds promising truth for precision oncology
This webpage was produced as an assignment for an undergraduate course at Davidson College.
Novel detection computer software accurately and effectively identifies cancer driver genes in rare cancers at the individual level
A key part of tackling cancer is understanding what makes an individual susceptible to developing the disease. Although an explosion of research has explored how particular environmental exposures such as tobacco use or UV radiation may contribute to the development of cancer, other genomic factors contribute to a fatal diagnosis (Parsa 2012). These external factors have been well established, but Nulsen and his colleagues further explore the significance of genetic components that contribute to esophageal adenocarcinoma, which exhibits high mutational and chromosomal instability leading to widespread genetic heterogeneity. 400 individuals with this cancer have been sequenced identifying numerous mutant and amplified genes (Mourikis 2019). Such a large number of potential recurrent driver events leaves too much ambiguity; therefore, the previous understanding of this cancer’s mechanisms were severely limited. Prior methods to identify such drivers heavily relied on comparison of alterations across patients. Although these had been of great value leading to the identification of more than 2,000 well established candidate driver genes, they failed to identify rare driver events that occurred in small cohorts or even in single patients because of low statistical power. Nulsen at al sought to subsequently investigate those cancers that could not be analyzed by these techniques with a recently developed patient level driver detection method: sysSVM2.
sysSVM2 at its most basic is an algorithm that uses machine learning with special emphasis on four support vector machines (Pisner 2020). This system can be used to identify driver alterations in patients belonging to rare cancer types for which assembling a large cohort of individuals is near impossible. The algorithm first works by identifying a set of true positive canonical drivers damaged within a cohort of patients and then ranking the remaining damaged genes. More highly ranked genes have higher similarities with canonical drivers and will be considered the cancer drivers for that patient. This algorithm was applied to 261 patients with esophageal adenocarcinoma. Nulsen et al studied 1,000 samples where 18,455 different genes were damaged 309,427 times. The assessed performance of sysSVM2 in predicting driver genes of 229 canonical drivers was significantly higher than that of any other patient level detection medium; the algorithm correctly identified the driver event in every patient. Therefore, sysSVM2 outperformed all others in both specificity and sensitivity. When compared to other driver detection methods, Nulsen et al’s system had better patient coverage and a particularly low rate of predicting established false positives, subsequently highlighting the future potential of sys SVM2.
Nulsen’s team posits that osteosarcoma served as a test in sysSVM2’s predictive capabilities because it is a rare, genetically heterogeneous bone cancer with poor prognosis and only 6 well established canonical drivers. Results showed that the sysSVM2 was able to identify reliable cancer driver genes across individual patients even for cancer types not outlined in training. This has relevant implications for obscure cancers that are poorly studied with little available genomic data. This example provided sound evidence confirming the suspected hypothesis that identifying the complete repertoire of driver events in each cancer patient holds great potential for furthering the molecular understanding of that cancer’s mechanisms. Once these mechanisms are well established and understood, potential treatments can be further explored and measured. That being said, such a system and a study design do not come without limitations. One being the fact there are several cancer genes that are not distinguished as a set of linked genes and hence they may be missed by this algorithm. Because of this short coming, researchers concluded that this approach ought to be used in conjunction with other already practiced methodologies. However, this study was successful in stably prioritizing canonical driver genes for most publicly available cancer cohorts. It can be used to identify driver alterations in individual patients and rare cancer types where canonical drivers are insufficient to explain the onset of disease, as validated with osteosarcoma. All of which furthers the field of precision oncology or “the molecular profiling of tumors to identify targetable alterations” as quintessential to mainstream clinical practice (Schwartzberg 2017).
Down the road, future research should explore the potential of such an individual level detection in other rare diseases that lack sufficient resources and data. Not only will this contribute to the growing field of oncology, it will provide another check on the effectiveness of sysSVM2. Potential diseases that Nulsen’s team could investigate next include leukemia, lymphoma, and thyroid cancers. All are rare and further investigation into their mechanisms and the inflicted genes could prove fruitful in decreasing the rates of progression and even preventing onset altogether. In addition to this, this algorithm could be manipulated to incorporate other analyses of the more inclusive technologies so that one day only one program may need to be used to analyze a set of patient data. This would eliminate the current limitation of sysSVM2. Taken together, this study was useful for answering the question of how to detect driver genes in rare cancers across individual genotypes. Not only did the use of sysSVM2 identify major players in such poor prognoses, it also provided a powerful tool by which future research can explore other rare cancers. All of this leaves the reader hopeful that precision oncology and this more personalized approach of medicine has the potential to both prevent and treat the devastating disease of cancer.
REFERENCES
- Parsa, N. Environmental factors inducing human cancers. Iran J Public Health. 41, 1-9 (2012). PMID: 23304670; PMCID: PMC3521879 (2012).
- Mourikis, T.P., Benedetti, L., Foxall, E. et al. Patient-specific cancer genes contribute to recurrently perturbed pathways and establish therapeutic vulnerabilities in esophageal adenocarcinoma. Nat Commun. 10, 3101 (2019). https://doi.org/10.1038/s41467-019-10898-3
- Pisner, D. A., & Schnyer, D. M. Support Vector Machine. 1, 1-20 (2020). https://www.sciencedirect.com/science/article/pii/B9780128157398000018
- Schwartzberg L, Kim ES, Liu D, Schrag D. Precision Oncology: Who, How, What, When, and When Not? Am Soc Clin Oncol Educ Book. 37, 160-169 (2017). https://ascopubs.org/doi/10.1200/EDBK_174176
© Copyright 2020 Department of Biology, Davidson College, Davidson, NC 38036.
Taylor J. Mingle is currently enrolled in Davidson College. Contact her at tamingle@davidson.edu.
