NEW YORK (GenomeWeb) – Genomenon, an Ann, Arbor, Michigan-based startup focused on developing tools for the personalized medicine space, is currently recruiting early adopters to test-drive Mastermind, a cloud-based curated database of genomic mutations, which is the company's first product for the marketplace.
Genomenon attended the annual meeting of the Association for Molecular Pathology held last week to showcase Mastermind and also to recruit additional early adopters, Mark Kiel, Genomenon's CEO and co-founder, told GenomeWeb. He said that so far the company has secured a few users for Mastermind but hopes to expand that initial group of testers who'll work with the company to refine and improve the database before the full launch in the first quarter of next year, he said.
When it launches, Genomenon will continue to offer the database under the cloud model although it could eventually offer an application programming interface that customers could use to access data, Kiel said. The company plans to charge a per-seat license for access to Mastermind and it is currently working out the exact pricing details.
Mastermind aggregates and captures information on variants, genes, and diseases from both structured and unstructured public and proprietary databases. In doing so it addresses what Genomenon believes is an important issue dogging personalized medicine efforts. Currently, information is siloed in disparate databases and geographic locations making it difficult for users to access and use, he said. "We have set about to compel order to the disparate data that exists for [next-generation sequencing-based] diagnostics ... by systematically combing the medical literature, looking for disease-gene mutation associations, and bringing a great deal of automation to that process."
Genomenon is populating Mastermind with the help of an automated querying infrastructure that it developed, which runs searches against existing variant and scientific literature repositories using input lists of diseases and all known genes and gene symbols as well as any known synonyms. The company's query engine combs available scientific resources for information on various cancers and cancer subtypes as well as genetic diseases.
The company's list of sources includes unstructured text data from sources such as the Online Mendelian Inheritance in Man database and ClinicalTrials.gov, as well as databases such as dbSNP, ClinVar, and the Catalogue of Somatic Mutations in Cancer. Furthermore, bespoke algorithms generate mutation query lists for each gene that the company identifies and those lists are used to search published primary literature including titles, abstracts, and supplementary information for any and all disease-gene mutation associations. These associations are prioritized in Mastermind based on the quantity and quality of the supporting pathogenicity information.
Mastermind users can query the database for information on variants of unknown significance or run searches based on diseases or genes. A query for information about a particular disease returns a list of all genes that are associated with that disease based on the literature. They can also input a list of genes or mutations into the database and get back a list of diseases that have been associated with those genes or mutations. For example, if a user searches for information on cherubism, a rare genetic disorder, Mastermind would return a list of about 12 mutations in the SH3BP2 gene based on information from about 450 articles on the condition that were mined by Mastermind's algorithms, Kiel said.
"Importantly, we show them the primary source material right down to the sentence from which we extracted that mutation from the article so that they can verify themselves that that association is a real one," he said. Users can then access the source material directly themselves if it's publicly available or go through appropriate channels for accessing private data.
Also, Mastermind is modular, so Genomenon can swap one data source for another if need be, he added. That includes new structured and unstructured databases that are both public and proprietary. "We can identify mutations in specific genes and whether or not they are associated with any diseases [in] that database, and interpolate those findings into the Mastermind database," said Kiel. For example, the company plans to add in pharmacogenomics information to the next version of Mastermind so that it can also provide drug-gene-mutation associations, Kiel said.
Savant and Prodigy
Genomenon was also at AMP last year showcasing Mastermind along with the two other products in its portfolio, Kiel said. Besides Mastermind, Genomenon also has developed two software tools that leverage the information contained in the database. The first of these is a tool called Savant that queries patients' mutation lists against Mastermind, prioritizes the variants based on several different clinical categories, and ranks the clinical categories based on the strength of the evidence that supports the mutation-disease link. The second, dubbed Prodigy, automates the identification of candidate biomarkers that contribute to disease in cohorts of patients with some biological similarity.
One example of Prodigy's use comes from Kiel's own research, in which he applied the tool to genomic sequences from 50 patients with T-cell prolymphocytic leukemia. Kiel explained that he used Prodigy to look for statistically valid genomic associations between these 50 patients such as which signaling pathways and biological functions were impacted in these patients. Prodigy prioritized mutations in the JAK/STAT signaling pathway, which Kiel and colleagues wrote in a paper published last year could offer opportunities for new targeted therapies.
Genomenon ultimately decided to launch Mastermind first because the database resonated the most with potential customers, Kiel told GenomeWeb. Many of these clients already had systems similar to Savant that they used internally for annotating important mutations, and as such were unlikely to abandon those pipelines for a new product. Since Mastermind addressed a current need in the marketplace and fit easily into existing pipelines, from a business perspective, it made the most sense to launch Mastermind first as a standalone product.
However, Genomenon believes that Savant takes a fundamentally different approach to analyzing genetic mutations that sets it apart from standard variant annotation and filtration programs and should earn it place in the market, Kiel said. In contrast to existing programs, Savant's emphasis is on prioritizing variants first.
The problem with the annotation approach, as Kiel sees it, is that the more annotations that are attached to patient's list of variants, the more information that the researcher or clinician has to look through and the greater the likelihood that they'll get false positives. A similar situation occurs with the filtration step, in which unannotated variants in the input list that may be significant for the patient in question are thrown out, resulting in false negatives.
Savant avoids these issues by focusing on identities and similarities between mutations in the patient's data and those in existing literature and databases, Kiel explained. As an example of how this works in Savant, a clinician might find information in the literature about a mutation in a given patient's genes that support its pathogenicity, but there could also be information about another mutation that hasn't been associated with a patient's condition in the same pathway or functional domain as the known pathogenic gene.
Access to that sort of information could be valuable to clinicians, Kiel noted. Returning to the aforementioned T-cell prolymphocytic leukemia cohort, he pointed out that the researchers found several palliative mutations across the 50 patients including one in interleukin-2 receptor gamma (IL2RG), which is upstream of the JAK/STAT signaling pathway. "If you have a patient who you suspect is having T-cell prolymphocytic leukemia and they have a mutation in IL2RG, which you've never seen before because those are rarer, it's not a valid approach to annotate and filter because you throw away that mutation, [even though] it is in the JAK/STAT signaling pathway and stays very close to the mutation that has been described."
Furthermore Savant comes with features such as direct access to read and coverage data, access to Mastermind content, and an interactive user interface, which should help make it attractive to potential customers, Kiel said. Genomenon plans to launch Savant in the summer of 2016. The company is also comparing Savant to both open-source and commercial packages and it plans to publish a white paper detailing its findings at a later date.
Meanwhile, the firm also has begun testing Prodigy with a select group of academic research laboratory clients, Kiel said. The company is currently providing biomarker discovery services and consulting based on the software to these users and charging a per-sample fee for analyses. However, the company plans to launch Prodigy as a standalone software product in spring 2016, Kiel said. The company is still working out exactly what it will charge. In addition to academia, Genomenon hopes this product will appeal to clients in the pharmaceutical industry.
Genomenon was officially incorporated in May 2014. It grew out of Kiel's post-doctoral research work at the University of Michigan on hematopoietic malignancies done in collaboration with Kojo Elenitoba-Johnson and Megan Lim, who are both cofounders of the company and also Genomenon's chief scientific officer and chief medical officer, respectively.
As part of efforts to analyze the genetic basis of these cancers, Kiel began using open-source software as well as developing his own bioinformatics tools and pipelines for processing and reporting the results of next-generation sequence analysis in clinical contexts. Those tools formed the bases for this first set of products that the company is working to commercialize in the coming months.
"We've got our sights on facilitating the promulgation of personalized medicine, [and] we foresee three problems with fully realizing the power of personalized medicine," Kiel told GenomeWeb. "There is the problem of the data, making discoveries efficiently and accurately, and then using both of those things to inform patient diagnoses. We've got three products ... that are meant to address each of those separate opportunities."
So far, Genomenon has raised about $600,000 in funding from grants through the University of Michigan and elsewhere, and from a number of competitions including a second-place prize it won in the Accelerate Michigan Innovation Competition, which added $100,000 to the company's coffers. In addition, Genomenon has raised $750,000 from angel investors.