Generating virtual perturbed cells utilizing a latent transfer perturbation neural network

A system comprising:

• at least one processor;
• and
• at least one non-transitory computer-readable storage medium storing instructions that, when executed by the at least one processor, cause the system to:
• generate, based on functional response data corresponding to phenomic or transcriptomic responses to cellular perturbations, cellular perturbation machine learning embeddings within a latent machine learning feature space comprising at least one of phenomic image machine learning embeddings or transcriptomic profile machine learning embeddings;
• determine pairwise similarity metrics among the cellular perturbation machine learning embeddings on alignment of the cellular perturbation machine learning embeddings within the latent machine learning feature space;
• generate a multi-perturbation interaction data object corresponding to relationships among the functional response data by constructing a unified representation of phenomic or transcriptomic interactions among the cellular perturbation machine learning embeddings utilizing the pairwise similarity metrics among the cellular perturbation machine learning embeddings;
• generate, utilizing a base state neural network, a control cell latent representation from features extracted from a control cell;
• generate, utilizing a perturbation neural network, a perturbation latent representation for a perturbation from the multi-perturbation interaction data object;
• generate a virtual perturbed cell latent representation by combining the control cell latent representation and the perturbation latent representation such that the virtual perturbed cell latent representation reflects phenomic or transcriptomic attributes of the perturbation from the multi-perturbation interaction data object applied to the control cell;
• and
• generate, utilizing a neural network decoder, a virtual perturbed cell response feature of a virtual cell subject to the perturbation from the virtual perturbed cell latent representation.

A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause a computing device to:

• generate, based on functional response data corresponding to phenomic or transcriptomic responses to cellular perturbations, cellular perturbation machine learning embeddings within a latent machine learning feature space comprising at least one of phenomic image machine learning embeddings or transcriptomic profile machine learning embeddings;
• determine pairwise similarity metrics among the cellular perturbation machine learning embeddings on alignment of the cellular perturbation machine learning embeddings within the latent machine learning feature space;
• generate a multi-perturbation interaction data object corresponding to relationships among the functional response data by constructing a unified representation of phenomic or transcriptomic interactions among the cellular perturbation machine learning embeddings utilizing the pairwise similarity metrics among the cellular perturbation machine learning embeddings;
• generate, utilizing a base state neural network, a control cell latent representation from features extracted from a control cell;
• generate, utilizing a perturbation neural network, a perturbation latent representation for a perturbation from the multi-perturbation interaction data object;
• generate a virtual perturbed cell latent representation by combining the control cell latent representation and the perturbation latent representation such that the virtual perturbed cell latent representation reflects phenomic or transcriptomic attributes of the perturbation from the multi-perturbation interaction data object applied to the control cell;
• and
• generate, utilizing a neural network decoder, a virtual perturbed cell response feature of a virtual cell subject to the perturbation from the virtual perturbed cell latent representation.