Download
In the discovery samples (WGS) in haploidentical HCT, totally we identified 2,259 LoF gene pairs associated with aGVHD (p < 0.01 and FDR < 0.1). To assess the combined effect of donor-recipient gene pairs, we taking the corresponding donor and recipient genes as covariates in logistic regression (aGVHD ~ LoF gene pair + LoF donor gene + LoF recipient gene). Among the 2,259 pairs, 666 are still significant in relation to aGVHD (LoF gene pair coefficient p < 0.05), indicating the important combined effects of LoF variations in pairwise donors and recipients.
Here, we provided the 666 pairs with the coefficient of LoF gene pair, as well as the coefficient p-value of the “LoF gene pair” terms.
With WGS sample set and exon targeted sequencing set of the consecutive cohort, we confirmed the aGVHD-related non-HLA variations including donor and recipient SNVs, mismatches, LoF genes as well as the LoF pairs. To systematically evaluate the effects of multiple risk factors on aGVHD onset, we built quantitative models combining confirmed non-HLA genetic variations, HLA type and matching degree as well as clinical characteristics such as donor and recipient sex, age, blood type, primary disease, and relationship.
Here, we provided the total feature list before selection. The detailed description of non-HLA paired genetic features are as follows:
The "paired" genetic features means genetic variations depend on both donor and recipient variations, including LoF pairs (http://agvhd.gao-lab.org/about.php#lofgenepair) and mismatches (different allele number of the SNP/inDel between donors and recipients). All these features for model selection are confirmed in both WGS set and target sequencing set (p<0.05 in the two sets). P is the coefficient p-value in logistic regression fitted to aGVHD. LoF gene paires are put into the candidate features which are significant in the following two scenario of regression:
1) aGVHD ~ LoF gene pair
2) aGVHD ~ LoF gene pair + LoF donor gene + LoF recipient gene
We also provided the confirmed the aGVHD-related non-HLA variations including donor and recipient SNVs, mismatches, LoF genes as well as the LoF pairs between WGS sample set (HLA-haploidentical HCT) and HLA-matched exon targeted sequencing cohort. Coupled with HLA and clinical characteristics, we used these features to build the HLA-matched model.
With the genome-wide screening, we identified novel genetic loci related aGVHD in terms of genotype-phenotype association, including donor/recipient/mismatched SNVs and structure variants (SV). These variants were not sequenced in the comfirmatory target gene panel for locating in intergenic or noncoding regions. Here, we provided the variants' location and significance level in this table (original Table S2).