US 12,288,600 B2
Generative machine learning on textual queries relating to molecules
Ben Sklaroff, San Francisco, CA (US)
Assigned to Genesis Therapeutics, Inc., Burlingame, CA (US)
Filed by Genesis Therapeutics, Inc., Burlingame, CA (US)
Filed on Jan. 9, 2024, as Appl. No. 18/408,042.
Claims priority of provisional application 63/510,599, filed on Jun. 27, 2023.
Claims priority of provisional application 63/460,421, filed on Apr. 19, 2023.
Claims priority of provisional application 63/437,832, filed on Jan. 9, 2023.
Prior Publication US 2024/0233883 A1, Jul. 11, 2024
Int. Cl. G16C 20/70 (2019.01); G16C 20/40 (2019.01); G16C 20/90 (2019.01)
CPC G16C 20/70 (2019.02) [G16C 20/40 (2019.02); G16C 20/90 (2019.02)] 20 Claims
OG exemplary drawing
 
1. A method performed by one or more computers, the method comprising:
presenting an interactive graphical user interface to a user on a display of a user device;
receiving, from the user and by way of the interactive graphical user interface, user input data defining: (i) a chemical structure of each of one or more input molecules, and (ii) a natural language textual query related to the one or more input molecules;
generating a sequence of input tokens that jointly represents: (i) the chemical structure of each input molecule, and (ii) the natural language textual query; and
processing the sequence of input tokens that jointly represents: (i) the chemical structure of each input molecule, and (ii) the natural language textual query, using a generative neural network to generate a sequence of output tokens defining data responsive to the natural language textual query, comprising:
sequentially generating each output token in the sequence of output tokens starting from a first output token in the sequence of output tokens;
wherein for one or more positions in the sequence of output tokens, generating the output token at the position comprises:
processing a network input that comprises: (a) the sequence of input tokens that jointly represents the chemical structure of each input molecule and the natural language textual query, and (b) a respective output token at each of one or more preceding positions in the sequence of output tokens, using the generative neural network, to generate a score distribution over a set of tokens; and
selecting the output token for the position in accordance with the score distribution; and
after generating the sequence of output tokens by, for each of one or more positions in the sequence of output tokens, selecting the output token for the position in accordance with the score distribution generated by the generative neural network:
presenting, to the user and by way of the interactive graphical user interface, a representation of the sequence of output tokens defining data responsive to the natural language textual query from the user,
wherein the representation of the sequence of output tokens comprises: (i) a sequence of output text, (ii) a respective visual rendering of a chemical structure of each of one or more output molecules, or (iii) both.