VAlue proposition
Deciphering the effects and mechanisms of genetic variants is essential to understand human genetics and further precision medicine. A genetic variant is a permanent change in the nucleotide sequence that makes up a gene, also known as a gene mutation. Mutations can cause diseases or phenotypic changes. Typically, Genome Wide Association Studies (GWAS) are used in genetic research to associate specific genetic variations with particular diseases. This method involves studying the genomes of a large group of people and searching for small variations of single nucleotide polymorphisms (SNIPS) then comparing the genome of a non-diseased organism. There are many issues with this method, as there are many individual associated variants unrelated to the disease in question. Recently, deep learning-based models have been developed to extract sequence information de novo to predict functional variant effects on nearby regions. Due to complex chromatin folding patterns and large-scale cell type specific data, it is difficult to decipher the effects of variants on nearby regions and understand their multilevel effects. There is a need for a comprehensive vision of genetic variants from genotype to phenotype considering complex chromatin folding.
Description of Technology
3DVarientVision is a deep learning based multi modal framework to provide a comprehensive vision of genetic variants effects from genotype to phenotype. 3DVarientVision integrates a diverse panel of genome-wide publicly available datasets such as ChIP-seq, Hi-C, and eQTL summary statistics. ChIP-seq combines chromatin immunoprecipitation assays with sequencing, identifying genome-wide DNA transcription factor and protein binding sites. Hi-C detects genome-wide chromatin interactions in the cell nucleus by utilizing 3C and next-generation sequencing. EQTL is a locus, explaining part of the genetic variance of a gene expression phenotype. A personal DNA sequence can be inputted, and 3DVarientVision pulls data from the aforementioned datasets to predict the genetic variants’ effects on specific TF bindings, histone modifications, enhancer-promoter interactions, and gene expression. 3DVarient further predicts probability of various diseases. 3DVarientVision accurately identifies the distal target genes of non-coding variants by capturing non-linear complex regulatory grammar across multiple levels of features and across long range genomic distances.
Benefits
Applications
IP Status
Copyright
LICENSING RIGHTS AVAILABLE
Full licensing rights available
Developer: Jianrong Wang, Jiaxin Yang
Tech ID: TEC2024-0091