Artificial intelligence technology has a bit of a bad reputation in computer science circles, but AI veteran Ben Goertzel is betting that the technology is just what the bioinformatics community needs.
Goertzel, who’s been an AI application developer since the 1980s, stumbled upon bioinformatics in 2001, and happily found that the problems that needed to be addressed in the field fit nicely with the type of work he was doing at the time, which involved integrating diverse types of data and recognizing subtle patterns in large data sets for the financial markets. Inspired by the potential of applying AI technology to biological data, he soon launched a company, called BioMind, with the goal of commercializing AI-based analysis of biological data.
At first, he said, “We were really motivated by a grand vision of AI and systems biology — trying to use artificial intelligence to integrate all known data about biological systems from molecules all the way up to whole organisms, and make a holistic picture of biological systems.” Soon recognizing that the scope of that initial vision was a bit out of reach, however, Goertzel said that BioMind opted to rein in its plans and tackle microarray analysis as a first step.
The company’s technology is built upon its BioMind AI Engine, which uses neural networks, formal logic, evolutionary programming, agent systems, and other AI approaches to integrate disparate data sets, while simultaneously inferring new knowledge and relationships from the information it is gathering. The AI Engine is used to create the BioMindDB database, which integrates around a dozen publicly available databases, including GenBank, PIR, the Gene Ontology, SGD, KEGG, Unigene, and MGI, as well as several gene expression datasets and chemical databases. Goertzel said that knowledge from BioMindDB is fully integrated within the algorithms that carry out analytical tasks such as clustering, supervised categorization, and time course analysis, in order to “recognize better patterns in the data.”
Other microarray data analysis products incorporate biological data into the analytical process, but according to BioMind, these products rely on the database knowledge to help with “the interpretation of analytical results, rather than directly to guide analysis.”
In terms of the categorization algorithms themselves, “We’ve done a lot of work to kind of tune and tweak machine-learning algorithms to work well with the knowledge from BioMindDB,” Goertzel said.
BioMind is currently working with the Centers for Disease Control and Prevention on a project to analyze gene expression data using its technology, “and we found some patterns that were of a great deal of interest to them,” said Goertzel. He added that he was unable to disclose further details because the results of the study have not yet been published. The company is also working with another customer in the area of Parkinson’s disease, Goertzel said.
BioMind’s technology has been under development for more than a year, and the company is rolling out its microarray analysis platform — BioMind Analyzer 1.0 — this quarter. Customers can either install BioMindDB and the AI Engine on their own servers, or opt for a hosting service that BioMind provides, said Goertzel. In the hosting model, “We set up separate servers for each customer, so there’s no problem with data security,” he said. In either case, the user interface is “any web browser.”
BioMind currently employs 15 people, which includes R&D, sales, and marketing. Goertzel’s first company, WebMind, closed in 2001 after a key investor canceled an expected multimillion-dollar round of funding, so this time around, he opted to support his company “the old-fashioned way”— through revenues.
Future plans for the technology include integrating an information-extraction feature within the AI Engine in order to add information from the biomedical literature into BioMindDB, Goertzel said. That feature is currently under development and will be released in 2004.
In addition, the company hasn’t abandoned its initial plans to focus on systems biology. On the contrary: Microarray analysis is the company’s “first stepping-stone” toward its long-term vision of “using AI-based data integration and data analysis to drive a rigorous systems biology,” Goertzel said.