PARIS--Genset’s CEO Pascal Brandys said this month that now is a “crucial time” for his company to aggressively exploit genomic data to create “multiple medically valuable assets.” Toward that end, the company recently raised over ¥49 million through issuance of convertible bonds exchangeable into Genset common shares. It has also, over the past year, established a formidable in-house bioinformatics presence.
In early 1999, Genset recruited Nicholas Schork to the position of vice president of statistical genomics. Schork, who has a PhD in epidemiology from the University of Michigan and master’s degrees in philosophy and statistics, gave up a position at Case Western Reserve University to head the French company’s sequencing analysis efforts from its La Jolla, Calif., laboratories.
Genset has boasted that it is attacking “statistical issues relevant to high resolution mapping and population-based statistics” using proprietary software in conjunction with its DNA sequence databases, NetGene and SignalTag. The company said its analysis software, dubbed BioIntelligence, utilizes “artificial intelligence tools” and “is developing rules for recognizing patterns characteristic of secretory proteins, genes, and regulatory regions that influence disease pathogenesis.”
Genset said its method for “linking” single-nucleotide polymorphisms with disease capitalizes on the fact that biallelic markers, which co-segregate with a disease-causing genetic variation but are not actually involved in the disease, are likely to occur at a spatial location near the genetic variation responsible for the disease. Once a SNP’s link to a disease is established, it can be used as a marker. The company is continuously updating and improving its map of markers.
While Genset officials wouldn’t say just how many SNPs it has found, it claims to have the highest quality SNP map in the world. Jim Kuo, vice president of worldwide business development, said, “The number of SNPs is much less relevant than the quality of SNPs mapped. For example, SNPs linked to genes can be useful while SNPs in nonsense sequence may be useless.”
This sophisticated SNP map is the core of the value the company offers its clients. In late 1998, to better take advantage of its SNP map, Genset acquired a massively parallel supercomputer “to support the computational needs generated by the exponential growth of its sequencing, mapping, and genotyping data.”
According to Schork, Genset differs from its competitors by being a collaborator rather than a service provider. “Genset’s staff and technologies are being applied in the planning and analysis of basic and clinical research for more than 10 different disease areas with seven different partners,” Schork said. The company focuses on association studies, functional genomics, and applied population genetics, he added.
Schork said he thinks Genset’s highly focused approach, although admittedly less industrial in scale than some of its competitors, is more likely to produce results in the near term. “Our studies are designed to find the impact of specific sequence information on the pathology of disease,” he said. “Both proprietary data and publicly available data are used to target specific sites in the genome for analysis in clinical studies and in model organisms such as knockout mice.”
Genset’s proprietary statistical tools and computer models determine probabilities for the functional impact of polymorphisms. Schork said his group is finding that the structure of genes and the variances that occur are much more complex than was thought.
“Our ability to use bioinformatics to acknowledge the complexity of genes and then design studies based on our findings is becoming more and more valuable as familiarity with the genome increases,” he said. For example, in designing association studies Schork pointed out that it is now well known that race is very arbitrary with respect to molecular variation. It is becoming desirable to be able to choose either homogeneous or heterogeneous genetic groups for trials in order to determine for what genetic groups of patients a potential drug will be effective.
Genset is secretive but enthusiastic about the properties of its analysis software. Jim Kuo said that the company is using the software in-house for proprietary research but added, “we do not care to publicly discuss the specific functions of the software until further extensions are completed.”
Kuo went on to say that the company is developing and patenting algorithms because it believes that “raw gene sequence information is becoming less and less valuable.” When questioned about the company’s ability to defend its proprietary algorithms from being used by others in the industry, Kuo pointed out that “there is precedent, mostly from the internet industry, for patenting and enforcing proprietary algorithms.” Kuo also mentioned that Genset has begun an academic microarray effort that will be combined with the bioinformatics effort.
Boguslaw Skierczynski, associate director of applied mathematics at Genset, described some of the company’s proprietary mathematical tools as “simply assessing probability density functions of certain events.” These tools, he claimed, allow Genset to identify linked markers more easily, thereby increasing the speed and reducing the cost of drug target identification. Although company representatives would neither confirm nor deny further probes regarding its proprietary analysis techniques, they did not deny the possibility that pattern recognition was being employed to search the genome based on known functional sequences found in secretory proteins and membrane bound receptors. Schork, Kuo, and Skierczynski all refused to comment on how artificial intelligence was being employed.