SAVANNAH, Ga.—Health Discovery Corp. may not hold all the cards when it comes to marketing and development of support vector machine (SVM) technology—the likely successor to neural networks. But it certainly has come very close with the recent completion of its acquisition of an extensive SVM patent portfolio previously owned by BioWulf Technologies and BioWulf Genomics.
Dr. Stephen D. Barnhill, the chairman and chief executive officer of Health Discovery, had been part of the BioWulf team working on SVM when he retired almost four years ago. But when BioWulf got into financial trouble, Barnhill became interested in the technology again as genomics and proteomics began to heat up and the idea of staying retired cooled down. Purchasing the patent portfolio of BioWulf was a natural next step, Barnhill says.
The patent portfolio is particularly valuable, Barnhill notes, because it includes several pending patent applications and issued patents—not the least of which are the pioneer patents in the use of the technology.
"There are hundreds of companies that have filed hundreds of patents in the SVM arena, and we hold the early priority dates on both issuance and filing of the original patents on SVM technology," says Barnhill. "The newer patents by others are more specific and narrow claims than what we hold, but many of them will probably have to license our patents to use theirs. I don't think that many people realize there are pioneer patents for SVM out there."
SVM is important because it is a technology expected to propel genomics and proteomics research and other arenas in which high-dimensional pattern recognition is required. While neural networks have provided a valuable technology base for pattern recognition thus far, Barnhill says, they are limited in how many inputs can be entered.
"The advent of high-throughput methods has revolutionized biomarker discovery and medical diagnosis but still poses major challenges to data analysis," Dr. Isabelle Guyon, a co-inventor of the SVM technology, said last year when Health Discovery Corp. was beginning the patent acquisition process. "Biologists are starting to acknowledge the power of machine learning, but few know how to harness this power."
SVM will change all of that, Barnhill says.
"I am an inventor on some of the early neural network patents in the 1990s," he says.
"I knew even then that as we forged into proteomics and genomics, we would need something better to handle inputs for 30,000 to 40,000 genes and hundreds of thousands of proteins—something that could handle the future of computational medicine."
SVM is that answer, he says, because it can handle pattern recognition in, essentially, infinite dimensional space.
Health Discovery's new patents are further bolstered by its acquisition last year of San Francisco-based Fractal Genomics and its patented bioinformatics software technology known as Fractal Genomics Modeling—which is able to find, link and model patterns of similarity hidden in large amounts of information, such as the clinical databases used for diagnostic and drug discovery.