This story has been updated to reflect inaccuracies regarding Qiagen's relationships with Millennium Science and Stanford University.
CHICAGO (GenomeWeb) – A month ago, Advaita Bioinformatics, an upstart maker of biomedical interpretation software, won a big deal in Australia and New Zealand, getting Millennium Science to distribute the US company's PathwayGuide and iVariantGuide technology for pathway and variant analysis.
Advaita had stepped in after Qiagen decided to end its relationship with Millennium Science to distribute Ingenuity Pathway Analysis in the Oceania region at the end of last year, according to a letter supplied by Qiagen. Qiagen said that it is now self-distributing the IPA product in Oceania.
It was not the first time Ann Arbor, Michigan-based Advaita had moved in on Qiagen's territory.
Stanford University's bioinformatics core, a heavy user of Qiagen IPA, also recently brought in iPathwayGuide by purchasing five seat licenses. Advaita President and CEO Sorin Draghici visited last month to train Stanford staff on the software.
"I showed them what the platform does. I thanked them for the invitation and I jumped in a cab to go to the airport," Draghici said. "By the time I got to the airport, I got a phone call from the home office that [Stanford had] ordered 10 more seats."
Stanford has since increased that order to a site license for the entire campus, said Draghici, who also is associate dean for innovation and entrepreneurship and director of the Anderson Engineering Ventures Institute at Wayne State University in Detroit.
A Qiagen spokesperson said that Stanford remains a customer. A paper published May 29 in Nature describes Stanford's use of IPA to analyze multi-omic data from patients with type 2 diabetes.
The technologies do differ somewhat. Rather than relying on gene-set enrichment analysis, iPathwayGuide uses a patented computational method for detecting and reducing pathway crosstalk. In April, Draghici and colleagues at Wayne State received a second patent for their computational method.
In terms of the pathway technology, competitors are typically looking at the number of differentially expressed genes between the phenotypes on a given pathway, according to Draghici. If the number is higher than the P-value, it is considered "significantly enriched," he said.
"This is very limiting because it ignores the very purpose for which these platforms have been created – the pathways, the background that shows you how the signals are transmitted from one gene to another, how one gene controls the next one, and so on," Draghici said.
That leaves out gene suppression and activation, as well as other signals, including the position of the gene on a pathway, he said
"If I have a pathway that has one gene as the key entry point, if that gene is knocked out and that gene is the only one with this information expressed, the pathway would be completely deactivated because that's the only entry point," Draghici said.
The Advaita mathematical model, described in a 2017 paper in Genome Research, considers the position and role of every gene on every pathway, then calculates the perturbation, according to Draghici.
"At every point in the pathway, you have a systems biology approach. We're using these advanced methods [and] artificial intelligence so we can better identify the pathways that are impacted and … whether the putative mechanisms would explain all the major changes throughout the system, taking into consideration all the measured changes," he said.
For its part, a few of Qiagen's analytics tools are similar to gene-set enrichment analysis, but most "would better be characterized as causal or predictive pathway analytics," CTO Ramon Felciano said via email.
"The majority of our customers buy IPA for the more sophisticated pathway-based algorithms that IPA has had for many years. If fueled with accurate and comprehensive pathway data, these algorithms can produce significantly more insightful results than gene-set-based approaches because they can leverage causal and probabilistic pathway information in ways that you cannot if you are only starting with gene sets," Feliciano said.
Advaita's technology builds on work from Draghici's Wayne State research lab. The company spun out of the university in 2005 and remained obscure for a while.
Advaita received a National Institutes of Health Phase I Small Business Innovation Grant of about $150,000 in 2008. The company has since received other grants from NIH and the National Science Foundation, making up the bulk of the $4.5 million Advaita has raised since its inception, according to Draghici.
The company currently makes three products. iPathwayGuide for proteomics and transcriptomics analysis; iVariantGuide for variant analysis; and the newest, iBioGuide, a bioinformatics search engine that Draghici likened to "Google for scientists."
In the burgeoning latter realm, Advaita has plenty of competition. Genomenon, IQvia's Linguamatics unit, and Saphetor all offer genomic search engines, as does French startup OmicX.
iBioGuide indexes information contained in iPathwayGuide and iBioGuide as well as things on the broader internet.
"If you Google and you do a search for breast cancer, you end up with Mayo Clinic definitions and treatment centers and support groups," Draghici said. "If you do the same on iBioGuide, you end up with all the genes that are related to that, all the variants, microRNAs, pathways, and so on, and you have all the information in one point."
That eliminates the need to jump around from website to website, including the Actionable Cancer Genome Initiative database to other databases specific variants and pathways. "It's in one place," Draghici said.
Advaita also has intellectual property in the discovery of subtypes of disease, Draghici said. While it is not part of the core platform yet, the firm has completed a custom integration of this technology with an unspecified pharmaceutical company.
A paper and a patent application in the works demonstrate how the company's technology can distinguish between patients with aggressive, life-threatening cancers and those with milder forms. Advaita actually uses "noise" in datasets to refine its AI engine, perturbing patient data with random noise to train the computer to find subtypes of cancers over and over.
"We then construct a matrix that shows which patients tend to end up in the same groups, regardless of the noise," Draghici said. "The main hypothesis here is that if true subtypes exist, the differences between those subtypes should be more important that the random differences between individuals."
This helps limit expensive, damaging treatments like chemotherapy and radiation in those with milder forms of cancer. "On the other hand, if you know the patient has a very aggressive form of disease, then you know that you have to go with the most aggressive treatment because that's their only chance," Draghici said.
He said that this type of predictive analytics can potentially save billions of dollars resulting from unnecessary treatments.