A company official incorrectly stated in an interview that the two sides of the company work together and reuse code developed on the services side for R&D efforts. That line has been removed from the article.
CHICAGO (GenomeWeb) – When it was spun out of drug discovery firm Selvita in 2015, Ardigen, a vendor of multi-omics artificial intelligence technologies for drug discovery and precision medicine, had a workforce of 20. Now, it employs about 80 people and this month moved into a new home office in Krakow, Poland to support its expansion.
Ardigen also is searching for a US headquarters, according to Director of Bioinformatics Marek Piatek.
The company launched with a focus on biological and clinical data analysis in genomics, transcriptomics, proteomics, metabolomics, and immunomics, and also had a business unit around personalized medicine products and services. It has since expanded into microbiomics, epigenomics, and analysis of CRISPR gene engineering data.
Ardigen started out making apps on a web platform, but Piatek said that the company is trying to evolve into a multi-omic software engineering expert. "We need to make sure that the data transfer and programming happens" according to customer needs and industry standards, he said.
The company sees big growth potential in immuno-oncology, microbiomics, and single-cell data services. "The microbiome and immuno-oncology, in those two legs of our R&D, we always support single-cell, RNA-seq analysis, all the variant calling and annotation," Piatek said.
Population-based precision medicine also represents a growth area.
"Some of the public [data] repositories are full of genomics, transcriptomics, and very often, epigenetics datasets. We are looking at what patterns [we can] find in the epigenomics datasets," Piatek said. "That can expand the phenotypes that we are observing."
Often, customers are starting from a phenotype and are looking at omics-related clues to find the cause of a condition, he noted.
Ardigen's R&D unit today mostly concentrates on immuno-oncology and the microbiomic effects on human health. This is where the company's technology development — bioinformatics, AI, and software engineering — takes place.
"There is a lot of analytics. We leverage the proprietary data of our customers together with the public-domain data," Piatek continued. Ardigen normalizes data, applies standards, and corrects for batch effects based on these datasets, and does not hold raw omics data in house.
Its services unit includes drug development portfolios. "We map our services onto each of the phases," Piatek said.
"We try to work in trios, if you will, where the software engineering would help us establish a safe and secure environment. We will make sure that the cloud is running properly, that the data transfer happens, and that the data scientists and bioinformaticians come in handy and solve biological problems," he said.
Customers include five of the 25 largest pharmaceutical companies, as well as some large academic medical centers, including Boston Children's Hospital and the affiliated Harvard Medical School, more on the research side than the clinical side of those health systems. Clients are mostly in North America and Western Europe.
Ardigen runs through a high-performance computing HPC cloud and is "platform-agnostic," according to Piatek. "We try to keep everything in a way that is not dependent on an environment."
He also said that this agnosticism extends to microbiome taxonomy. While other bioinformatics vendors might rely on taxonomy-related strain identification, Ardigen looks for genomic features in microbiomes.
"A genomic feature might exist in multiple strains, but what we care about is the genomic feature itself," Piatek said. "At the moment, we're not interested in which bacteria it belongs to."
In immuno-oncology, Ardigen likes to analyze information about the presentation and binding affinity of peptides, he said.
Immuno-oncology does underpin much of the company's work.
"We try to apply all our knowledge around cancer in general," Piatek said. "We're looking for driver mutations, we're looking for passenger mutations, tumor mutational burden, antigen burden, and all the things that are related to cancer and tumors."
But he noted that because Ardigen does include so many different types of omics, it can support research not necessarily related to cancer and not necessarily related to drug discovery.
Data from marketing literature suggested that Ardigen's predictive modeling technology outperformed MoleculeNet, a Stanford-developed large-scale benchmark for molecular machine learning. While MoleculeNet was described in a 2018 Chemical Science article, Ardigen has not yet published its findings in a peer-reviewed journal, though that is in the works.
"We have a chemo-informatics team that is heavily engaged with our laboratory partners that actually validates some of the machine learning experiments that we are running in silico. They are actually confirming and validating this in vitro now," Piatek said. That will be published soon, he said.
As a bioinformatics software developer, Ardigen does take part in open-source forums. It has published one repository on Github, a model for molecular optimization called Mol-CycleGAN.