US 11,783,915 B2
Convolutional neural network systems and methods for data classification
Virgil Nicula, Cupertino, CA (US); Anton Valouev, Palo Alto, CA (US); Darya Filippova, Sunnyvale, CA (US); Matthew H. Larson, San Francisco, CA (US); M. Cyrus Maher, San Mateo, CA (US); Monica Portela dos Santos Pimentel, San Jose, CA (US); Robert Abe Paine Calef, Redwood City, CA (US); and Collin Melton, Menlo Park, CA (US)
Assigned to GRAIL, LLC, Menlo Park, CA (US)
Filed by GRAIL, LLC, Menlo Park, CA (US)
Filed on Sep. 29, 2022, as Appl. No. 17/936,529.
Application 17/936,529 is a continuation of application No. 16/428,575, filed on May 31, 2019, granted, now 11,482,303.
Claims priority of provisional application 62/679,746, filed on Jun. 1, 2018.
Prior Publication US 2023/0045925 A1, Feb. 16, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G16B 30/10 (2019.01); G16H 50/20 (2018.01); G16B 40/30 (2019.01); G16B 40/20 (2019.01); G06N 3/04 (2023.01); G06N 3/084 (2023.01)
CPC G16B 30/10 (2019.02) [G06N 3/04 (2013.01); G06N 3/084 (2013.01); G16B 40/20 (2019.02); G16B 40/30 (2019.02); G16H 50/20 (2018.01)] 40 Claims
OG exemplary drawing
 
1. A computer system for classifying a cancer condition, in a plurality of different cancer conditions, the computer system comprising:
at least one processor;
a graphical processing unit having a graphical processing memory configured to store a network architecture; and
a memory, the memory storing at least one program for execution by the at least one processor, the at least one program comprising instructions for:
(A) obtaining, for each respective training subject in a plurality of training subjects of a species: (i) a cancer condition and (ii) a genotypic data construct that includes genotypic information corresponding to locations of a reference genome of the species, thereby obtaining a plurality of genotypic data constructs;
(B) formatting each genotypic data construct in the plurality of genotypic data constructs into a corresponding vector set comprising one or more corresponding vectors, thereby creating a plurality of vector sets;
(C) providing the plurality of vector sets to the network architecture that includes at least (i) a first neural network path comprising a first plurality of layers including at least a first network layer associated with at least a first filter comprising a first set of filter weights and (ii) a scorer, wherein the first network layer sequentially receives vector sets in the plurality of vector sets;
(D) obtaining a plurality of scores from the scorer, wherein each score in the plurality of scores corresponds to an input of one of the vector sets in the plurality of vector sets into the network architecture; and
(E) using a comparison of respective scores in the plurality of scores to the corresponding cancer condition of the corresponding training subject in the plurality of training subjects to adjust at least the first set of filter weights thereby training the network architecture to classify a cancer condition, in the plurality of cancer conditions.