Tudor I Oprea¹

Graham S Timmins²

¹ Dept. Internal Medicine, School of Medicine, University of New Mexico

² Dept. Pharmaceutical Sciences, College of Pharmacy, University of New Mexico

Many biomedical researchers discover -the hard way-that obviousness, as defined by the US patent and trademark office (USPTO) greatly influences patent allowance. A finding of obviousness prevents allowance, and courts and USPTO carefully construct and refine its definition. The recent development of Inter Partes Review (IPR) allows rapid and inexpensive challenge of issued patent claims, with most petitioners using arguments based around obviousness.  Such patent invalidation could have enormous implications to brand name drug makers, given that the global pharmaceutical market nears 3 trillion USD (2011-2015). Therefore, obviousness can cause major financial hits to biotechnology and small molecule drug firms, as well as many other areas of commerce.

Linguistically, obviousness means “easily perceived or understood; apparent”, stemming from Latin, obvius, from obviam, “in the way, within reach”.  So how does USPTO define obviousness? Well, it’s  definition is complex and with a lot of guidelines, but reductio ad absurdum involves performing this thought experiment:  1) we use a time machine to go back in time to the invention date with the patent examiner or IPR panel (because it would be very easy to find something was obvious using hindsight); and we 2) find someone who had an ordinary level of skill in the field of this invention and who knew all of the previously-known information (prior art) surrounding the invention (let’s assume we can all agree on an ordinary level of skill, although we probably won’t); and we 3) ask this person “ Is this invention obvious to you? “ (Although we’d all use different guidance to get the answers we wanted). Not that there’s anything wrong with thought experiments, they worked relatively well for Albert Einstein, but, could there be another way of doing an ‘obviousness check’ to see whether our thought experiment had given a fair result?

The idea was to look back at what a panel of ‘peer inventors’ were doing in the field of similar inventions before and after the time of the invention, based upon examining what they actually disclose in their patent applications, and when they did so. If the entire panel of inventors (say the top 20 Pharmaceutical companies) had similar ideas (of say, combining certain things) well before the date of the invention being examined, one could argue it was obvious to them, and this might apply to all. Equally, if you had filed a patent before most of this panel of peer inventors when they had all the motivation in the world to do so, then perhaps things weren’t quite as obvious as they might have seemed from a conventional analysis.

Until recently, it would take a lot of effort to analyze thousands of patent applications using text searching, but developments in artificial intelligence and text mining mean that sophisticated and nuanced searches are now readily performed.  Furthermore, at the American Chemical Society national meeting in San Diego a couple of weeks ago, Google announced that chemical (sub)structure searching is now enabled across all of Google Patents (which includes Google Scholar). This is due to advances in automated text mining strategies and text-based chemical structure recognition, enabled by name-to-structure technologies from OntoChem and RDKit. This represents a powerful additional tool for mining data relevant to obviousness.  

For example, the approval date (12/17/96) and patent expiration date (8/11/18) for atorvastatin, as mined from DrugCentral, suggest that patents were filed before 1997.  Using the chemical structure (and similar) to search Google Patents shows the earliest patents were in 1988-1992, with Warner-Lambert, later acquired by Pfizer, having a 2/22/88 priority date.  After 2/22/88, the first 11 patents are also assigned to Warner-Lambert.  Thus, we can safely state that atorvastatin was invented, that its synthesis was not obvious, and that Warner-Lambert’s intellectual property was consolidated with more than 10 patents filed after the initial one. 

Since patent databases are impartially maintained by agencies such as USPTO the playing field is transparent, and if one was required to disclose the search strategies and peer panel make up as well as final results, the fairness or otherwise of the analysis could be ‘cross examined’ just as easily.  We look forward to seeing how these ideas are (hopefully) used to help determining what is, and is not, obvious.