Non-linear correlation analysis method based on joint structure constraint and incomplete multi-modal data

Abstract

A non-linear correlation analysis method based on joint structure constraint and incomplete multi-modal data obtains multiple modal phenotype data and weights corresponding to SNP through four steps, non-linear correlation between SNP and phenotypes is constructed through non-linear transformation, and therefore complex correlation between SNP and phenotypes is considered; and through contribution of a plurality of SNPs to phenotypes, modal sharing and modal specific biomarkers corresponding to different modalities are obtained. The minimum root-mean-square error is obviously superior to the minimum root-mean-square error value obtained by the prior art, so that the performance of detecting the biomarker can be improved.

Description

technical field [0001] The invention relates to the field of application technology based on genetic data structure information and its incomplete multimodal data, in particular to a nonlinear association analysis method based on joint structural constraints and incomplete multimodal data. Background technique [0002] Du et al. (L.Du et al., "Multi-Task Sparse Canonical Correlation Analysis with Application to Multi-Modal Brain Imaging Genetics," IEEE / ACM Transactions on Computational Biology and Bioinformatics, vol.18, no.1, pp.227- 239, 2021.) proposed a multi-task sparse canonical correlation analysis (MTSCCA) method to identify disease-associated SNPs and multimodal phenotypes using multimodal image data generated by different imaging techniques that may carry complementary information. At the same time, the method takes into account the structural correlation between genetic data and the sparsity of genetic and phenotypic data at the individual level. By taking this i...

AI Technical Summary

Problems solved by technology

[0003] However, there are some problems in the MTSCCA method. First, due to imaging quality and high cost, most of the multimodal image phenotype data have missing data. This method removes the missing part of the sample and only applies the complete multimodal image. data for modeling, which may lose some information, thereby degrading detection performance

Second, the MTSCCA method only pays attention to the characteristic information of a single modality, but does not take into account the correlation between inter-modal and intra-modal data

Third, the MTSCCA method uses a linear model to identify the linear association between SNPs and phenotypes. However, the association between SNPs and phenotypes is very complicated, and it is difficult to detect such a complex relationship with a simple linear model.

Images