Three-way parallel independent component analysis for imaging genetics using multi-objective optimization

摘要: In the biomedical field, current technology allows for collection of multiple data modalities from same subject. consequence, there is an increasing interest methods to analyze multi-modal sets. Methods based on independent component analysis have proven be effective in jointly analyzing modalities, including brain imaging and genetic data. This paper describes a new algorithm, three-way parallel (3pICA), identifying genomic loci associated with function structure. The proposed algorithm relies use multi-objective optimization identify correlations among maximally sources within modality. We test robustness approach by varying effect size, cross-modality correlation, noise level, dimensionality Simulation results suggest that 3p-ICA robust SNR levels 0 10 dB effect-sizes 3, while presenting its best performance high correlations, more than one subject per 1,000 variables. experimental study 112 human subjects, method identified links between (pointing mental disorder genes, PPP3CC, KCNQ5, CYP7B1), functional related signal decreases default mode network during task, structure indicating increases gray matter regions region. Although such findings need further replication, simulation in-vivo validate ICA presented here as useful tool decomposition applications.