Graph Embedding Method Based Genetical Trajectory Reveals Migration History Among East Asians

Abstract

An important issue in human population genetics is the ancestry. By extracting the ancestral information retained in the single nucleotide polymorphism (SNP) of genomic DNA, the history of migration and reproduction of the population can be reconstructed. Since the SNP data of population are multidimensional, their dimensionality reduction can demonstrate their potential internal connections. In this study, the graph and structure learning based Graph Embedding method commonly used in single cell mRNA sequencing was applied to human population genetics research to decrease the data dimension. As a result, the human population trajectory of East Asia based on 1000 Genomes Project was reconstructed to discover the inseparable relationship between the Chinese population and other East Asian populations. These results are visualized from various ancestry calculators such as E11 and K12B. Finally, the unique SNPs along the psudotime of trajectory were found by differential analysis. Bioprocess enrichment analysis was also used to reveal that the genes of these SNPs may be related to neurological diseases. These results will lay the data foundation for precision medicine.East Asia is one of the birthplaces of human culture (Su et al., 1999). Especially the ancient Yellow River Basin civilization and the large-scale agricultural society form fostered by the Yangtze River Basin have laid an irreplaceable material and cultural foundation for the development of civilization in East Asia. The early Chinese cultural model has greatly influenced the cultural patterns of the Japanese islands, the Korean peninsula and the Vietnamese peninsula, as well as parts of the Southeast Asia mainland. The vocabulary derived from classical Chinese has profoundly changed the language systems of these regions. However, the genetic landscape of this area, as well as the early migration of humans in East Asia, are still unclear. Based on historical records and information provided by archaeological literature, we can learn many clues that can be used for limited understanding, but due to historical records and non-scientific, almost "mythological" descriptions, many recorded "legends" seem to account for bigger. Some Asian human paleontology papers published in recent years have given very important information in this field (Su et al., 1999), but the dynamic migration process of the overall population structure is still very vague. Therefore, it is necessary to explore population migration patterns based on internal connections of population.Our research used the combination of the principle graph and structure learning based Graph Embedding method and tSNE to reduce data dimension. Based on the R package "Monocle" we first filtered the mutations, then reduced the data dimension, and calculated trajectory with psudotime and states; we also adopted all the equilibrium mutations to tSNE method firstly to reduce the dimension, and then uses DDRTree to reduce the dimension twice. The E11 calculator and the K12B calculator verify the sampling res...