Influence Analysis Of Weight Selection In Variant Set Mixed Model Association Tests For Continuous And Binary Traits

Genome wide association studies aim to find the variant loci that are significantly associated with diseases or traits in the whole human genome.Fruitful achievements have been made in the early stage,but studies are mainly carried out on common variants(minor allele frequency MAF>5%).In recent years,researchers have begun to focus on the effects of rare variants on diseases or traits.The methods of rare variant association test are mainly divided into single variant test and variant set test,and variant set test is usually more efficient.It is noted that population stratification and relatedness are the main confounding factors in the association test of rare variant,and the weight of single variant in the variant set test is usually required.Therefore,this dissertation focuses on the selection of weight in the variant set mixed model association tests(SMMAT).Firstly,Four test methods of SMMAT are introduced for continuous and binary characters.Secondly,the Beta probability density function weighting method based on MAF,Beta weighting method,is introduced.Afterwards,a score statistic for each rare variant is derived under the condition that the single rare variant is not associated with the trait.Drawing on the method of combining P value of exact test,combining with Beta weighting method,a new rare variation weighting method,SVA-Beta weighting method,is proposed.Finally,this dissertation compares the Type 1 error rates and power of the above two weight methods in SMMAT through simulation studies,and applies them to the actual data analysis of the iOmics dataset.From the simulation results,the Type Ⅰ error rates of the SMMAT test under the two weighting methods are reasonable.For power,the conclusions were as follows:(1)for SMMAT-B,SMMAT-S and SMMAT-E tests,power increased with the increase of the proportion of causal variation,and with the increase of the proportion of negative effect from 50%to 100%;For continuous traits,when the proportion of causal variation was low or the proportion of negative effect was 50%,SVA-Beta weight method had higher power.For binary traits,SVA-Beta weight method has higher power and better adaptability to data.(2)For SMMAT-S test,no matter continuous or binary traits,the test power of the two weight methods is not ideal.In comparison,the power of Beta weight method is lower.The results of iOmics actual data analysis showed that the SMMAT test under two weighting methods could detect rare variation sets significantly associated with SM(35:1)lipids.