NEW YORK (GenomeWeb) – By combining whole-genome sequencing and artificial intelligence-based analysis, researchers have reported that they were able to quickly identify clinically actionable mutations within a brain tumor sample.
Researchers from the New York Genome Center, Rockefeller University, and IBM analyzed a glioblastoma sample through panel testing as well as whole-genome sequencing. That sequencing data was then analyzed by a team of bioinformaticians and oncologists at the center as well as by IBM's Watson for Genomics.
As they reported in Neurology: Genetics yesterday, the researchers found that sequencing uncovered more clinically actionable mutations than panel testing and that relying on Watson for Genomics rather than human analysis reduced the time it took to identify those mutations.
"This study documents the strong potential of Watson for Genomics to help clinicians scale precision oncology more broadly," Vanessa Michelini, study co-author from IBM Watson Health, said in a statement. "Clinical and research leaders in cancer genomics are making tremendous progress in the opportunity to bring precision medicine to more cancer patients, but genomic data interpretation is a significant obstacle, and that's where Watson can help."
For their study, the researchers collected tumor and normal samples from a 76-year-old man with glioblastoma. At initial resection, a sample from the patient underwent analysis with Foundation Medicine's FoundationOne test.
Meanwhile, the researchers also performed whole-genome sequencing on the tumor and normal samples. Using the Illumina HiSeq X platform, they sequenced the tumor sample to more than 75X coverage and the normal sample to 42X coverage to uncover nearly 8,500 SNVs and 431 indels.
When bioinformaticians and oncologists assessed the findings, they uncovered six actionable SNVs, including ones in MET, FGFR3, and PIK3R1. Of the five CNVs they found, two affected genes also had SNVs.
In particular, the researchers found both a deletion and an amplification involving MET that appeared to lead to MET overexpression and exon-skipping, as gauged by RNA-seq.
At the same time, they noted an insertion in PIK3R1, which interacts with and inhibits PIK3CA. However, they reported that this PIK3R1 variant binds PIK3CA, but doesn't inhibit it, suggesting it leads to PIK3CA activation.
Together, the NYGC analysis suggested combining a MET and a PIK3CA inhibitor to treat this patient and noted a clinical trial that was assessing the combination. The patient's condition, however, deteriorated before he could enter the trial, and he died eight months after resection.
By contrast, the researchers reported that the FoundationOne assay uncovered fewer actionable targets and fewer potential treatments. Sequencing, they said, uncovered eight variants the panel test did not.
The researchers also used a beta version of Watson for Genomics to analyze the sequencing data. Watson for Genomics processes abstracts and, when possible, full-text articles from PubMed as well as the trials listed at ClinicalTrials.gov to identify cancer-linked variants and treatments that target them.
In this study, Watson for Genomics also uncovered six potentially actionable alterations, five of which overlapped with the NYGC analysis.
Similar to the manual analysis, Watson for Genomics recommended BKM120 as a potential therapeutic option targeting the overexpression of PIK3R1 and noted two drugs — crizotinib and cabozantinib — that could target the MET amplification.
Watson for Genomics also provided results more quickly. While human analysis of the patient's VCF file took 160 hours, the researchers reported that Watson delivered results in 10 minutes. "This is critical if sequencing is to be brought out of the research arena and into the scaled, real-world clinical realm," the researchers wrote in their paper.