Periodic Reporting for period 1 - AI4SYN (Artificial intelligence for synthetic functional genomics of blood)
Reporting period: 2022-09-01 to 2025-02-28
Every day, billions of blood cells need to be replaced in our bodies. This remarkable feat is achieved by the blood-forming system, where a small number of stem cells continuously gives rise to progenitor cells that massively proliferate and generate all blood and immune cells. Hematological diseases, such as leukemias, myelodysplasias and cytopenias, have their origin in a malfunction of that system. Conversely, stem cell transplants can be used to cure many genetic diseases of the blood- and immune system, such as sickle cell disease. Together, the blood forming system is a fascinating model for cellular differentiation, and of high biomedical relevance.
Stem cell differentiation is controlled by genes and gene regulatory networks. The activation of genes defines cellular identity; for example, a red blood cell needs to make hemoglobin alongside other proteins required for its function. Gene regulatory networks (GRNs) govern the activation of genes. Specifically, GRNs are composed of transcription factors and gene regulatory elements such as enhancers. Transcription factors are proteins that bind to gene regulatory elements and regulated the activation or repression of nearby genes.
A precise map of the GRN that drives blood stem cell differentiation, alongside a quantitative understanding of its function, would be an invaluable asset. It would, for example, allow to predict interventions to overcome differentiation blocks in leukemia, or to design enhancers that drive the specific activation of gene therapies in cell types of interest.
The overall objective of AI4SYN is to achieve such an understanding. To that end, AI4SYN generates systematic perturbation data of gene regulatory networks as well as data on the function of fully synthetic enhancers; these data allow to chart the blood GRN qualitatively, and are ideally suited for training AI models that quantitatively predict the effect of interventions to the GRN, or the function of enhancers.
Stem cell differentiation is controlled by genes and gene regulatory networks. The activation of genes defines cellular identity; for example, a red blood cell needs to make hemoglobin alongside other proteins required for its function. Gene regulatory networks (GRNs) govern the activation of genes. Specifically, GRNs are composed of transcription factors and gene regulatory elements such as enhancers. Transcription factors are proteins that bind to gene regulatory elements and regulated the activation or repression of nearby genes.
A precise map of the GRN that drives blood stem cell differentiation, alongside a quantitative understanding of its function, would be an invaluable asset. It would, for example, allow to predict interventions to overcome differentiation blocks in leukemia, or to design enhancers that drive the specific activation of gene therapies in cell types of interest.
The overall objective of AI4SYN is to achieve such an understanding. To that end, AI4SYN generates systematic perturbation data of gene regulatory networks as well as data on the function of fully synthetic enhancers; these data allow to chart the blood GRN qualitatively, and are ideally suited for training AI models that quantitatively predict the effect of interventions to the GRN, or the function of enhancers.
AI4SYN has generated systematic, functional data on systematically designed enhancers, as well as on the effects of systematic perturbations of the GRN driving hematopoiesis. Thereby, we have obtained a detailed, qualitative understanding of the function of key transcription factors and their interactions, both in the context of regulatory DNA, and in the context of the wider gene regulatory network. For example, we could show that transcription factors of the blood forming system often repress each other, thereby making sure that lineage genes are repressed in stem cells, where these factors are co-expressed. The systematic nature of our data has allowed us to train AI models that enable the design of new enhancers with cell state specificity.
Our results enable the design of completely artificial cell state specific enhancers from scratch, with potential applications in gene therapy