|[part of a heat map of gene expression values; from Wikimedia]|
expression profiles or genes (independent claim 10). Illustrative claim 1 reads:
1. A method for determining similarity between portions of gene expression profiles in a computer comprising the steps of:
- processing a number of gene expression profiles with a similar sequences algorithm that is a time and intensity invariant correlation function to obtain a data set of gene expression profile pairs and a match fraction for each gene expression profile pair;
- listing gene expression profile pairs in clusters by their match fractions;
- removing a first gene expression profile from a cluster when another cluster has another gene expression profile with a higher match fraction with the first gene expression profile, unless the another gene expression profile requires a larger number of subsequences to achieve similarity with the first gene expression profile;
- repeating the removing step until all gene expression profiles are listed in only one cluster
- providing output of the listing of clusters of gene expression profiles.
The assesment of patentability by the Board of Appeals was based on the following observations:
- The method according to claim 1 is performed "in a computer";
- The specification teaches that "[t]he present invention may also be used in any digital computer architectures, including personal, minicomputer and mainframe computer environments, and in local area and wide area computer networks";
- Claim 1 "does not transform any physical object or article";
- The specification teaches that a preferred embodiment "uses a time and intensity-invariant correlation function such as that described by R. Agrawal. [...] Specifically, we employ the similar sequence algorithm embodiment of the above described correlation function in Intelligent Miner for Data [...] which was designed for business intelligence, against time varying gene expression data."
In a second examination stage, the Board then analysed claim 1 under the Supreme Court’s precedents on the unpatentability of abstract ideas (see earlier posting). It found that "claim 1 applies the mathematical manipulation of data to determine the similarity between portions of gene expression profiles" (cf. observation 4), such that “the application’s only innovation is its reliance on a mathematical algorithm" (referring to Parker v. Flook and Bilski v. Kappos) making claim 1 an abstract idea as applied to defined data.
Regarding program product claim 10, the Board concluded that the expression “stored on a recordable medium”, when interpreted in the broadest reasonable way may read on the non-transitory "carrier wave storage" of In re Nuijten (for explanation see here or here) and referred to the USPTO's latest guiding on "Subject Matter Eligibility of Computer Readable Media" (see this earlier posting).
The Board thus took the position that, for different reasons, both claim 1 and claim 10 fail to define statutory subject matter as required under 35 U.S.C. § 101.
Criticism: In the meantime the BPAI decision has received some attention from US patent bloggers. While Karen G. Hazzah of All Things Pros is concerned with the rejection of claim 10 and felt that the "Board [not only] should have explained (at a minimum) just exactly why 'carrier wave storage' is 'transitory' and thus non-statutory under In re Nuijten" but also interpreted In re Nuijten wrongly, Steve Lundberg of Patents4Software criticises the Board's reasoning with respect to method claim 1.
He argues that claim 1, firstly, satisfies both the machine prong (limited to a computer implementation) and the transformation prong (transformation of data items representing biological properties) of the machine-or-transformation test and, secondly, does not claim an 'abstract idea' but rather is directed to a "very specific task of processing data representative of gene expression, which itself is representative of physical properties of biological material" and "does not merely calculate a 'number' but rather produces 'clusters of gene expression profiles' that are not mathematical in nature, but rather reorganized or correlated data representative of biological properties of biological material".