A more complete haplotype prediction is still a mathematical calculation, and storing and providing it adds nothing

2021-03-11Fed. Cir.

In re Board of Trustees of the Leland Stanford Junior University, 989 F.3d 1367, No. 2020-1012 (Fed. Cir. Mar. 11, 2021) (precedential) (Reyna, J., joined by Prost, C.J., and Lourie, J.), affirming the Patent Trial and Appeal Board's rejection of claims 1, 4 through 11, 14 through 25 and 27 through 30 of United States Patent Application No. 13/445,925

United States Patent Application No. 13/445,925, filed by Stanford on April 13, 2012, claims a method for resolving haplotype phase, the determination of which parent an allele was inherited from. Claim 1 recites receiving allele data for a family of two parents and at least two children, receiving pedigree data, determining an inheritance state using a hidden Markov model with specified hidden states, receiving transition probability data and population linkage disequilibrium data, determining a haplotype phase from those inputs, storing it, and providing it in response to a request, each step on a computer system comprising a processor and a memory. The specification reports phase resolution of 97.9 percent of heterozygous positions against roughly 80 percent for the prior method. The examiner rejected the claims under Section 101 of Title 35 of the United States Code, the Board affirmed, and the Federal Circuit affirmed. At step one claim 1 is directed to the abstract idea of mathematically calculating alleles' haplotype phase, recites no concrete application for the haplotype phase beyond storing it and providing it upon request, and a greater number of predictions is not an improved technological process. At step two the hardware was generic, the mathematical steps and data types conventional according to the specification itself, and the ordered combination unexplained. Dependent claims adding diagnosis, drug treatment and prognosis determinations instruct only to apply the algorithm.

What it changesThe decision is the Federal Circuit's clearest statement yet that a better inference, standing alone, is not a patent-eligible invention. The claim recited a specific model with named hidden states and specific input data, and that specificity did not matter because the output of the method was a stored number. For anyone drafting machine learning claims, the independent claim needs a step that does something with the inference that is not storing it, displaying it or sending it on request: controlling equipment, altering a physical process, or changing how the computer itself operates. The dependent claims here show what does not work, because a generic step of determining a drug for treatment from the output was treated as apply it. Where the engineering work is an improved prediction and nothing downstream is claimable, the application should be reconsidered against the trade secret alternative before the filing budget is spent, and the specification should at least describe the computational improvement in terms of the machine rather than the result.

More on Eligibility of software and AI claims

← All recent developments · Search all developments

Burbage Law is the practice of Jon Burbage, a registered patent attorney. 342 United States patents have issued from applications he wrote; the full list is on the Patents page. More about Jon Burbage.

This is general information about a published decision, agency guidance or piece of legislation. It is not legal advice and does not create an attorney-client relationship. Items concerning offices outside the United States are reported for information only and are not counsel on the law of those countries.