‘This is amazing’: New tool can help assess clinical impact of somatic cancer mutations
Click Here to Manage Email Alerts
Next-generation sequencing and precision medicine techniques have enabled the identification of millions of somatic cancer variants.
Although this information has the potential to greatly improve targeted cancer care, the classification and interpretation of these somatic variants is a challenging and time-consuming aspect of tumor genetic profiling, according to Marilyn M. Li, MD, professor of pathology and laboratory medicine and director of cancer genomic diagnostics at Children’s Hospital of Philadelphia (CHOP).
“There are two steps in cancer genomic data analysis that are currently most time-consuming — one is variant classification and the other is interpretation and reporting,” Li, who has collaborated in the development of a new tool to streamline this process, told Healio. “This tool will not only reduce the time and manual work involved, but will also increase the consistency of our variant classification.”
Li and colleagues published a paper characterizing the tool, CancerVar, in Science Advances. She spoke with Healio about CancerVar, which incorporates machine-learning technology to interpret the potential significance of somatic mutations in terms of cancer diagnosis, prognosis and treatment, including targeted therapy.
Healio: Why is the classification and interpretation of somatic variants so time-consuming?
Li: Cancer is not like other diseases where you have one mutation associated with one disease. With cancer, it’s not only multiple mutations, it’s multiple types of mutation of many different genes, which is why normally, we don’t just sequence one gene. We use next-generation sequencing technology to sequence hundreds of genes and numerous mutations at the same time, so that generates thousands of variants. We have to figure out what is clinically significant and what is normal polymorphism or artifact. For those that are cancer-related, we have to determine the driver and the passenger mutation. You want to focus on those driver mutations that can inform the patient’s care. The CancerVar tool will save a good deal of time in this process.
Healio: What does the CancerVar tool provide in terms of classification and interpretation?
Li: The CancerVar tool is unique in the way it sources multiple currently available data sets. It has compiled 13 million cancer-associated mutations from 1,911 cancer census genes that were mined through existing studies and databases. CancerVar efficiently captures quality big somatic cancer mutation data and utilizes artificial intelligence (AI) tools to categorize them based on the somatic variant interpretation guidelines. The CancerVar tool helps not only to classify the variants, but also sources whatever publications are available to give you the clinical and laboratory experimental evidence to support your variant classification. It also provides the evidence of clinical significance for reporting these variants. This is extremely important in the clinical laboratories that provide genomic testing for precision cancer care.
Healio: Has the CancerVar tool shown efficacy in classifying and interpreting somatic variants?
Li: Absolutely. When you use machine-learning technology to inform your classification, the bigger the data and the better the quality of the data, the better the accuracy and consistency. You don’t have to do as much manual work as without AI support. In terms of interpretation, this software will give you the data you need for interpretation and reporting, which is a tremendous help. If you can automate all the steps, this can not only reduce the time spent, but it also makes your test cheaper. I truly believe we can and should make our tests cheaper and better, so we can bring the benefit afforded by the new technologies to all pediatric patients with cancer.
Healio: When will this be introduced into clinical practice? Can it be used on adult patients, as well as pediatric patients?
Li: We have done some validation in the cases we have written about in the development of the software. Some of our manual, human-evaluated variants were validated with this pipeline. At CHOP, we have multiple different tests for cancer and we will combine all of them into one large capture and use the newer machine with larger capacity for next-generation sequencing. Once we get that new test validated, we are going to implement and integrate CancerVar into our pipeline. This technology will be perfect for adult cancer cases, as well. The databases this pipeline has sourced and integrated are not just for pediatric cancers; there are more databases for adult patients than for pediatric patients. At CHOP, we also get cases from adult hospitals.
Healio: Is there anything else you’d like to mention on this topic?
Li: CancerVar can automate the clinical practice guidelines, and this is amazing. I think this is, basically, the first time in the cancer setting that we can automate the guidelines and the classification and interpretation process. However, this is by no means to completely replace all manual work; rather, it will reduce a significant amount of manual work. I anticipate that in the next few years, we will see a lot of people want to incorporate this or similar tools into their bioinformatic pipelines.
References:
- Li MM, et al. J Mol. Diagn. 2017;doi:10.1016/j.jmoldx.2016.10.002.
- Li Q, et al. Sci Adv. 2022;doi:10.1126/sciadv.abj1624.
For more information:
Marilyn M. Li, MD, can be reached at Children’s Hospital of Philadelphia, 3615 Civic Center Blvd., RM716-ARC Bldg., Philadelphia, PA 19104; email: lim5@chop.edu.
Click Here to Manage Email Alerts