US 12,142,350 B2
Method for predict affinity between drug and target substance
Yeachan Kim, Gyeonggi-do (KR); and Bonggun Shin, Gyeongsangbuk-do (KR)
Assigned to DEARGEN INC., Daejeon (KR)
Appl. No. 18/558,164
Filed by DEARGEN INC., Daejeon (KR)
PCT Filed Mar. 15, 2022, PCT No. PCT/KR2022/003582
§ 371(c)(1), (2) Date Oct. 30, 2023,
PCT Pub. No. WO2023/033281, PCT Pub. Date Mar. 9, 2023.
Claims priority of application No. 10-2021-0115509 (KR), filed on Aug. 31, 2021.
Prior Publication US 2024/0266006 A1, Aug. 8, 2024
Int. Cl. G16C 20/30 (2019.01); G16C 20/40 (2019.01)
CPC G16C 20/30 (2019.02) [G16C 20/40 (2019.02)] 11 Claims
OG exemplary drawing
 
1. A method for predicting an affinity between a drug and a target substance, the method performed by a computing device including at least one processor, the method comprising:
extracting a feature value of each of the drug and the target substance by using a first neural network;
performing a cross attention between the feature values that considers information of the target substance to align the feature value of the drug and the feature value of the target substance by using a second neural network; and
predicting the affinity between the drug and the target substance based on a result of performing the cross attention and based on the calculated feature value of the target substance for the drug and the calculated feature value of the drug for the target substance by using a third neural network, wherein the affinity represents a binding force or force acting between the drug and the target substance;
wherein the performing of the cross attention between the feature values by using the second neural network includes:
generating a first key, a first query, and a first value for the drug based on the feature value of the drug, and generating a second key, a second query, and a second value for the target substance based on the feature value of the target substance;
generating a cross attention vector based on the second query and the first key, and calculating the feature value of the target substance for the drug using the first neural network by applying the cross attention vector to the first value; and
generating a second cross attention vector based on the first query and the second key, and calculating the feature value of the drug for the target substance using the first neural network by applying the second cross attention vector to the second value.