US 11,657,898 B2
Biological interaction and disease target predictions for compounds
Maria Chatzou, London (GB); Pablo Preto Barja, London (GB); Niklas Henrik Benjamin Kokkola, London (GB); and Rishabh Shukla, London (GB)
Assigned to LIFEBIT BIOTECH LIMITED
Filed by LIFEBIT BIOTECH LIMITED, London (GB)
Filed on Feb. 4, 2021, as Appl. No. 17/167,944.
Application 17/167,944 is a continuation of application No. PCT/EP2020/059686, filed on Apr. 3, 2020.
Claims priority of application No. 1904887 (GB), filed on Apr. 5, 2019.
Prior Publication US 2021/0158897 A1, May 27, 2021
Int. Cl. G16B 25/10 (2019.01); G16B 40/30 (2019.01); G16B 5/20 (2019.01); G16B 40/20 (2019.01)
CPC G16B 25/10 (2019.02) [G16B 5/20 (2019.02); G16B 40/20 (2019.02); G16B 40/30 (2019.02)] 18 Claims
 
1. A computer implemented method comprising:
receiving, at a cell digital twin, input data corresponding to data previously acquired from a biological cell line previously exposed to a compound of interest;
generating, by the cell digital twin, predictions about biological interactions between the compound of interest and biological cells based on the input data, wherein:
the cell digital twin comprises a prediction engine, the prediction engine comprising a generative neural network trained using known cell response profiles corresponding to tested compounds, and
the prediction engine generates biological interaction predictions between the compound of interest and the biological cells using the generative neural network, the biological interaction predictions including one or more of predicted immune response, predicted immune coverage, predicted mode of action, or predicted drug action and target diseases;
generating, by a validation engine, validated predictions using the biological interaction predictions generated by the cell digital twin, wherein:
the validation engine generates validation information by parsing one or more images of one or more references, processing text of the one or more references, and searching the processed text and the parsed images of the one or more references for one or more of the predicted immune response, predicted immune coverage, predicted mode of action, or predicted drug action and target diseases, and
the validated predictions include the biological interaction predictions and the validation information; and
updating the prediction engine by providing the validated predictions to the prediction engine using a feedback loop between the validation engine and the prediction engine.