CHICAGO (GenomeWeb) – Sophia Genetics has moved into liquid biopsies with an update to its artificial intelligence system.
The Swiss analytics company unveiled the new app for liquid biopsies that runs on SOPHiA —the AI technology embedded in the Sophia Data Driven Medicine (DDM) platform — at the American Society for Clinical Oncology meeting here this month. It followed the introduction recently of the firm's Whole Exome Solution and Clinical Exome Solution for the detection of disease-related genetic variants.
This new app interprets circulating tumor DNA and circulating tumor cells in blood, urine, cerebral spinal fluid and other liquid samples to help clinicians make earlier diagnoses and better treatment decisions, as well as to monitor the efficacy of cancer treatments without having to wait months to see changes in follow-up imaging scans. The app also supports clinical trial recruitment, according to Sophia Genetics, which is based in Saint-Sulpice, Switzerland, near Lausanne.
CEO and cofounder Jurgi Camblong called SOPHiA "more sophisticated" than earlier genomic informatics systems from Sophia Genetics and others.
"This is like cycling or running," Camblong said. "Before running, you have to be able to walk."
 The walking part means addressing "simple needs" like compensating for sequencing mistakes and inaccuracies.
"Once you're good at that, you can move into the more complex issues," such as degraded DNA and, eventually, ctDNA, Camblong said. This is happening now.
SOPHiA AI draws on a deep database the company has been building since development on DDM began in 2012 and since the platform hit the market in 2014. Sophia Genetics has a customer network of about 305 hospitals in more than 50 countries, and clients add more than 8,000 analyses per month to the database, according to Camblong.
This, Camblong said, creates a "diversity of data," leading to a high level of accuracy for the artificial intelligence. "When you think about artificial intelligence, you think data. When you think data, you think network," he explained.
SOPHiA presents results of its analyses through OncoPortal, an interface that Sophia Genetics launched in October 2016 to help match patients' tumor profiles to actionable information on available treatments and clinical trials. The company said that OncoPortal helps clinicians interpret AI outputs by flagging known associations between gene variants, disease causes and progression, drug efficacy, and treatment side effects.
Camblong said the technology is accurate with first-variant detection, and that the precision is increasing as volume builds.
"We will get to know the tumor of a patient" as the database gets larger and larger, Camblong said.
"You can determine who will benefit from a drug" before writing a prescription for what typically is an expensive treatment, he explained. But that takes refinement, which is an ongoing process.
Last year, Camblong told GenomeWeb that internal, pre-release testing found that SOPHiA predictions of pathogenic variants in BRCA genes matched those of clinical experts 85 percent of the time when the algorithms were trained with data from 10,000 patients. Predictive accuracy increased to 96 percent with data from 20,000 patient tests and climbed to 98 percent with data from 30,000 patients and 28,000 unique variants identified in their tests.
The new technology is immediately available via the cloud to any hospital that wants it. "That, for us, is the magic here." The target market is every hospital doing next-generation sequencing, he said.
An early adopter of SOPHiA for liquid biopsies is Hospices Civils de Lyon, the second-oldest teaching hospital in France. In January, Hospices Civils de Lyon became the first research organization to try this new SOPHiA application.
"It is easy to carry out a molecular profile for patients," said Léa Payen-Gay, a biologist at Hospice Civils de Lyon.
Sébastien Couraud, a physician in the hospital's Department of Specialized Acute Pulmonology and Thoracic Oncology, said the AI technology has helped find "small, relevant models" in some cancers. For example, Lyon researchers located mutations that could serve as predictive biomarkers to help clinicians choose therapies.
"Liquid biopsy is a new paradigm," Couraud said. "We can detect mutations from these small fragments [of ctDNA]."
Looking for ctDNA makes detecting multiple biomarkers easier. "For me, the clinician, I can have in one blood sample several different analyses [over time] and adapt treatments," Couroud said.
Payen-Gay said she uses SOPHiA for analyzing samples in the hospital after running a sequence.
Initially, her team takes a full biomarker panel, with which clinicians can monitor individual biomarkers as patients progress and see the efficiency of treatments over time. This reduces stress on patients and can lower costs by avoiding invasive biopsies and inappropriate drug choices.
Couraud said the hospital has been adding about 40 new samples to the AI platform every month, which leads to "clinical-grade information" for clinicians to act on.
Plus, a liquid biopsy tends to be more accurate than a tissue biopsy, since the latter is just a sample from one area and not a full representation of a disease or treatment efficacy, Couraud noted. Blood samples tend to have clones of mutations, so clinicians can get a more comprehensive picture of the spread or remission of a cancer.
"Tissue is not still the issue," Couraud said.