ABSTRACT
Mapping gene networks requires large amounts of transcriptomic data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues. Recently, transfer learning has revolutionized fields such as natural language understanding1,2 and computer vision3 by leveraging deep learning models pretrained on large-scale general datasets that can then be fine-tuned towards a vast array of downstream tasks with limited task-specific data. Here, we developed a context-aware, attention-based deep learning model, Geneformer, pretrained on a large-scale corpus of about 30 million single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology. During pretraining, Geneformer gained a fundamental understanding of network dynamics, encoding network hierarchy in the attention weights of the model in a completely self-supervised manner. Fine-tuning towards a diverse panel of downstream tasks relevant to chromatin and network dynamics using limited task-specific data demonstrated that Geneformer consistently boosted predictive accuracy. Applied to disease modelling with limited patient data, Geneformer identified candidate therapeutic targets for cardiomyopathy. Overall, Geneformer represents a pretrained deep learning model from which fine-tuning towards a broad range of downstream applications can be pursued to accelerate discovery of key network regulators and candidate therapeutic targets.
Subject(s)
Biology , Machine Learning , Neural Networks, Computer , Humans , Biology/methods , Single-Cell Gene Expression Analysis , Datasets as Topic , Chromatin/genetics , Chromatin/metabolism , Cardiomyopathies/drug therapy , Cardiomyopathies/genetics , Cardiomyopathies/metabolismABSTRACT
Retinitis pigmentosa (RP) is the most common form of inherited vision loss and is characterized by degeneration of retinal photoreceptor cells and the retinal pigment epithelium (RPE). Mutations in pre-mRNA processing factor 31 (PRPF31) cause dominant RP via haploinsufficiency with incomplete penetrance. There is good evidence that the diverse severity of this disease is a result of differing levels of expression of the wild-type allele among patients. Thus, we hypothesize that PRPF31-related RP will be amenable to treatment by adeno-associated virus (AAV)-mediated gene augmentation therapy. To test this hypothesis, we used induced pluripotent stem cells (iPSCs) with mutations in PRPF31 and differentiated them into RPE cells. The mutant PRPF31 iPSC-RPE cells recapitulate the cellular phenotype associated with the PRPF31 pathology, including defective cell structure, diminished phagocytic function, defects in ciliogenesis, and compromised barrier function. Treatment of the mutant PRPF31 iPSC-RPE cells with AAV-PRPF31 restored normal phagocytosis and cilia formation, and it partially restored structure and barrier function. These results suggest that AAV-based gene therapy targeting RPE cells holds therapeutic promise for patients with PRPF31-related RP.