- From: John Wilbanks <wilbanks@w3.org>
- Date: Mon, 12 Jul 2004 11:40:05 -0400
- To: public-semweb-lifesci@w3.org
One of the points of discussion at the pharmagrid retreat last week pricked my ears and I wanted to throw it out here for discussion. It revolved around workflows, and how intellectual property theory is starting to gel around them. The concept of a "workflow" in the pharma community is not a new one. Here's an example...a very basic one. 1. After running gene expression analysis, get 54,000 data points representing 18,000 genes (three spots per gene for significance testing) back from robot. 2. Test significance and toss out genes with too much variation amongst the three. 3. Normalize the data according to your preferences (or prejudices, as it were) 4. Choose algorithm and distance metric for clustering. Now, the impact of semantics on life sciences means that a second workflow could be applied to the clusters coming out of step 4 above. Such as. 1. Represent list of clusters as text list of gene aliases with semantic associations to activity: "on/off" (and if "on" then "upregulated/downregulated" associations attached) 2. Filter for genes that make proteins that are nuclear proteins 3. Filter again for genes that make proteins that are involved in cell death 4. Filter yet again for genes that make proteins that have a molecular signature similar to my candidate drug compound 5. Filter yet again for genes that make proteins that have a "minimal" representation in literature (i.e., less likely to have restrictive intellectual property around them) 6. Rank according to my ontology of the subclasses of semantic constraints in 2-5 and the on/off, upregulated/downregulated relationships, because some nuclear proteins are more equal than others in pharma. Ones that create the biological activity you want when they are turned "off" are ideal, since it's easier to silence a gene without side effect than it is to activate one without side effect. And so forth. On two different occasions at the conference I heard mention of "protecting" this second type of workflow as a business process representative of the skill of the designer of the workflow. In one case it's already happened, though the office action is not known. This strikes me as a place where the concept of a precompetitive repository of workflows might make a lot of sense. Anyone out there have an opinion? jtw
Received on Monday, 12 July 2004 11:41:09 UTC