scGen uses a variational autoencoder model to project gene expression measurements into a latent space, where a vector δ is obtained that represents the difference between perturbed and unperturbed cells from the training set.
This project will develop and demonstrate a combined experimental/computational systems biology approach for identifying effective antibody properties for host immune defense against pathogen infection and then determine appropriate microenvironment conditions for influencing B lymphocytes to generate antibodies exhibiting these properties. This will be accomplished through 2 tasks pursued in parallel. Task 1: Develop and apply integrative multi-variate experimental and computational methods enabling mathematical association of protective antibody features with corresponding B cell transcriptomic profiles. Task 2: Develop and apply a computational framework for inference of B cell regulatory network states from transcriptomic profiles and determine how desired network states may be attained by appropriate cytokine and/or small molecule treatments.
We are envisioning future work beyond the scope of this foundational project to connect these in vitro culture insights to in vivo contexts.