Mutual information adversarial autoencoder

1. A method for generating an object, the method comprising:

providing a dataset having object data for an object and condition data for a condition;

processing the object data of the dataset to obtain latent object data and latent object-condition data with an object encoder;

processing the condition data of the dataset to obtain latent condition data and latent condition-object data with a condition encoder;

processing the latent object data and the latent object-condition data to obtain generated object data with an object decoder;

processing the latent condition data and latent condition-object data to obtain generated condition data with a condition decoder;

comparing the latent object-condition data to the latent condition-object data to determine a difference;

processing the latent object data and latent condition data and one of the latent object-condition data or latent condition-object data with a discriminator to obtain a discriminator value;

obtaining better latent object data, latent object-condition data or latent condition-object data, and/or latent condition data based on the discriminator value, difference between the latent object-condition data and latent condition-object data, and difference between object with generated object and condition with generated condition;

selecting a selected object based on the generated latent object data and latent condition-object data from a given condition;

obtaining a physical form of the selected object; and

validating the physical form of the selected object.

2. The method of claim 1 , comprising performing:

comparing the generated object data with the object data; and

selecting a selected generated object data that is less than a threshold object difference between the generated object data and the object data.

3. The method of claim 2 , comprising performing:

comparing the generated condition data with the condition data; and

selecting a selected generated condition data that is less than a threshold condition difference between the generated condition data and the condition data.

4. The method of claim 3 , comprising selecting the selected object that corresponds with the selected generated object data or that corresponds with the selected generated condition data.

5. The method of claim 1 , further comprising:

preparing the physical form of the selected object; and

testing the physical object with the condition.

6. The method of claim 5 , wherein:

the obtaining of the physical form of the selected object includes at least one of synthesizing, purchasing, extracting, refining, deriving, or otherwise obtaining the physical object; and/or

the testing includes assaying the physical form of the selected object in a cell culture; and/or

assaying the physical form of the selected object by genotyping, transcriptome-typing, 3-D mapping, ligand-receptor docking, before and after perturbations, initial state analysis, final state analysis, or combinations thereof.

7. The method of claim 1 , wherein:

the processing with the object encoder and the processing with the condition encoder are performed simultaneously and independently; and/or

the processing with the object decoder and the processing with the condition decoder are performed simultaneously and independently.

8. The method of claim 1 , wherein the difference is a latent deviation loss L MSE of an Euclidean distance between the latent object-condition data and latent condition-object data.

9. The method of claim 8 , wherein the discriminator value is in a range of [0,1], and a loss function of the difference between the discriminator value and 1 is determined as the L adv .

10. The method of claim 9 , further comprising:

defining the latent object data as L object and defining a weighting of the latent object data as W object ;

defining the latent condition data as L condition and defining a weighting of the latent condition data as W condition ;

defining a weighting of the L MSE as W MSE ;

defining a weighting of the L adv as W adv ;

calculating a weighted sum as an L total ; and

determining whether L total is below a threshold, when the L total is below the threshold, the selected object is selected,

wherein:

L total =W object *L object +W condition *L condition +W MSE *L MSE +W adv *L adv .

11. The method of claim 1 , wherein the latent object-condition data is substantially equal to the latent condition-object data.

12. The method of claim 1 , comprising:

assuming the object is complex; and

assuming the condition is complex.

13. The method of claim 1 , wherein the dataset is a pairs dataset of pairs (x,y) wherein the object is x and the condition is y, the pairs dataset including:

z x\y as a variable corresponding to data specific for x and thereby for the object;

z y\x as a variable corresponding to data specific for y and thereby for the condition; and

z x∩y as a variable corresponding to data common between x and y and thereby common between the object and the condition.

14. The method of claim 13 , wherein the dataset includes data for molecule-protein binding,

z x\y includes data for one or more molecules that bind with a target protein;

z y\x includes data for different molecules with similar binding properties to the target protein; and

z x∩y includes data linking one or more specific molecules to one or more specific binding sites of a target protein.

15. The method of claim 13 , wherein the dataset includes data for molecules, cell states prior to interacting with a molecule, and cell states subsequent to interacting with the molecule;

z x∩y includes data that describe molecule-cell interaction;

z x\y includes data on variation of molecules that cause the same change in a cell transcriptome; and

z y\x includes data for a transcriptome that is not related to the molecule-induced changes.

16. The method of claim 1 , wherein the selected object is a molecule, the obtaining of the physical form includes synthesizing the molecule, and the validating includes:

obtaining transcriptome data of a cell prior to interacting with the molecule; and

obtaining transcriptome data of the cell subsequent to interacting with the molecule.

17. The method of claim 16 , further comprising determining whether the molecule satisfies the condition.

18. The method of claim 17 , comprising:

determining the molecule is similar to one or more molecules in the object data; and

determining the molecule has an activity similar to the one or more molecules in the condition data.

19. The method of claim 18 , comprising determining the molecule is distinct from other molecules with other mechanisms of action.

20. A computer program product comprising:

a non-transient, tangible memory device having computer-executable instructions that when executed by a processor, cause performance of a method comprising:

providing a dataset having object data for an object and condition data for a condition;

processing the object data of the dataset to obtain latent object data and latent object-condition data with an object encoder;

processing the condition data of the dataset to obtain latent condition data and latent condition-object data with a condition encoder;

processing the latent object data and the latent object-condition data to obtain generated object data with an object decoder;

processing the latent condition data and latent condition-object data to obtain generated condition data with a condition decoder;

comparing the latent object-condition data to the latent condition-object data to determine a difference;

processing the latent object data and latent condition data and one of the latent object-condition data or latent condition-object data with a discriminator to obtain a discriminator value;

obtaining better latent object data, latent object-condition data or latent condition-object data, and/or latent condition data based on the discriminator value, difference between the latent object-condition data and latent condition-object data, and difference between object with generated object and condition with generated condition;

selecting a selected object based on the generated object data and latent condition-object data from a given condition; and

providing the selected object in a report with a recommendation for validation of a physical form of the object.