By Allison Proffitt
January 24, 2019 | Genomic data interpretation is a crowded field. But how do you know that a platform works? How do you know that the solution you’ve picked is really the right one?
Ali Torkamani compares a potential solution to tools he’s created in the past.
Torkamani, Director of Genomics and Genome Informatics at Scripps Research Institute, built a genomics analysis platform himself once; the platform was acquired by Human Longevity. He knows first-hand what it takes to build a good tool.
“It’s a lot of work, basically! It’s a combination of keeping the pipelines updated as new databases come online and making sure you have the processes in place to keep things up to date. A lot of little things, really, you have to take into account to keep pushing your accuracy up more and more.”
In his role at Scripps, Torkamani says, “It just didn’t make sense to try to continue doing that internally.”
Six months ago, Torkamani began using the Genoox genomic analysis platform. He’s been validating the platform by comparing Genoox analytics with his lab’s previous findings for about 70 trios or family groups from molecular autopsy and rare disease work. In nearly every case, the Genoox findings have aligned with the lab’s previous findings. Twice Genoox has returned a diagnosis for cases for which Torkamani’s team had not been able to find results.
Torkamani has used his previous experience developing a solution to evaluate the Genoox platform.
“I’m familiar with the various pitfalls that arise when you’re doing genomic interpretation because I’ve built something like that before. I know what to look out for. I can ask them whether or not they’re looking out for these pitfalls,” he says. “Genoox does a good job addressing the pitfalls that I know of, and they’re also very responsive when we raise other concerns.”
Play It Again, Sam
Reanalysis with Genoox solved two of Torkamani’s cases, and the company is counting on that being a regular occurrence moving forward.
Today, Genoox announces artificial intelligence-based tools for regular and periodic genomic data reanalysis and data aggregation to quicken diagnoses for patients based on new clinical evidence, and improve diagnostic yield for undiagnosed cases.
“One of the things you have today in genetics, in general, is that in some of the cases where you have millions of mutations per genome, you need to go over them, score them, and understand what they mean. In many of the cases you’re still left with unknown variants—VUSs,” Amir Trabelsi, CEO and founder of Genoox tells Diagnostics World. “Today there’s no signal for it, but what if some publication is published two days after you do your analysis? What if you have a new patient that comes to your clinic that provides some more evidence or signals?”
Reanalysis—especially an automated reanalysis—could reveal new diagnoses or give greater context for a variant of unknown pathogenicity.
“Reanalysis, itself, is not yet regulated; an organization doesn’t have to do reanalysis,” Trabelsi says. “But I believe that we will see some regulations around this in the next few months or so.”
If Trabelsi is right, Genoox will be ready.
The Genoox platform already combines multidimensional data—family history, clinical data, genomics, and data from similar patients—and crawls public and proprietary databases to provide relevant insights. Now the platform will systematically reanalyze data and alert users if something has changed—either from a non-actionable finding to an actionable finding, or if a previous finding has been reclassified as benign.
Timing and data inputs are up to the users. “We can be highly customizable and granular, but we usually give a defined template to not confuse customers with so many parameters,” Trabelsi explained.
“It’s a matter of how strict or lose you want the reanalysis protocol to be,” he says. Some users will choose to only use trusted and known repositories of data, no research evidence. Others will want to be alerted to all changes on a mutation.
For strictly clinical applications Trabelsi expects reanalysis two to three times a year, he says. And many groups will limit findings to signal changes according to the guidelines. “Our solution will tell you not only if the data has changed, but also if the classification of the data has changed according to the guidelines, for example ACMG or NCCN,” he says.
Torkamani hasn’t used the Genoox’s new AI capabilities yet, but he doesn’t mince words when reporting his impression of Genoox so far. He praises the platform’s user interface, the company’s responsiveness, and the comprehensiveness of the solution.
“Genoox just makes our work much more efficient,” he said, “so we can focus on the research, essentially, and try to impact people’s lives.”