Irfahan Kassam

Rare coding variants shape inter-individual differences in human phenotypes¹. However, the contribution of rare non-coding variants to those differences remains poorly characterized. Here we analyse whole-genome sequence (WGS) data from 347,630 individuals with European ancestry in the UK Biobank^2,3 to quantify the relative contribution of 40 million single-nucleotide and short indel variants (with a minor allele frequency (MAF) larger than 0.01%) to the heritability of 34 complex traits and diseases. On average across phenotypes, we find that WGS captures approximately 88% of the pedigree-based narrow sense heritability: that is, 20% from rare variants (MAF < 1%) and 68% from common variants (MAF ≥ 1%). We show that coding and non-coding genetic variants account for 21% and 79% of the rare-variant WGS-based heritability, respectively. We identified 15 traits with no significant difference between WGS-based and pedigree-based heritability estimates, suggesting their heritability is fully accounted for by WGS data. Finally, we performed genome-wide association analyses of all 34 phenotypes and, overall, identified 11,243 common-variant associations and 886 rare-variant associations. Altogether, our study provides high-precision estimates of rare-variant heritability, explains the heritability of many phenotypes and demonstrates for lipid traits that more than 25% of rare-variant heritability can be mapped to specific loci using fewer than 500,000 fully sequenced genomes.

Validation and characterization of stromal DEGs with scRNA-seq and immune cell deconvolution analysis. A, Dotplot showing the percentage of stromal (nonmalignant) cells expressing each gene for each response category. Color represents the average scaled expression of the gene in stromal cells (PD, progressive disease; SD, stable disease; PR, partial response). B, Cell-type specific expression of the top stromal biomarkers in ICB responders (PR) and nonresponders (PD/SD). C, Heat map of the association between immune cell types and ICB response. Abundance of immune cell types of each tumor is estimated using CIBERSORTx. Heat map is colored by the signed −log10(P-value) of the difference in CIBERSORTx scores between responders (MSI) and nonresponders (MSS). Cell types with no significant difference in abundance between ICB responders and nonresponders (P-value >0.1) were colored gray. D, Consensus correlations between the abundance of immune cell types and expression of the top stromal biomarkers. Mean Pearson correlation coefficients across the seven discovery cohorts were plotted for each cell type-gene pair. Positive correlations were colored red, negative correlations were colored blue, and tiles with Pearson correlation coefficient <0.1 were colored gray.

Validation and characterization of stromal DEGs with scRNA-seq and immune cell deconvolution analysis. A, Dotplot showing the percentage of stromal (nonmalignant) cells expressing each gene for each response category. Color represents the average scaled expression of the gene in stromal cells (PD, progressive disease; SD, stable disease; PR, partial response). B, Cell-type specific expression of the top stromal biomarkers in ICB responders (PR) and nonresponders (PD/SD). C, Heat map of the association between immune cell types and ICB response. Abundance of immune cell types of each tumor is estimated using CIBERSORTx. Heat map is colored by the signed −log10(P-value) of the difference in CIBERSORTx scores between responders (MSI) and nonresponders (MSS). Cell types with no significant difference in abundance between ICB responders and nonresponders (P-value >0.1) were colored gray. D, Consensus correlations between the abundance of immune cell types and expression of the top stromal biomarkers. Mean Pearson correlation coefficients across the seven discovery cohorts were plotted for each cell type-gene pair. Positive correlations were colored red, negative correlations were colored blue, and tiles with Pearson correlation coefficient <0.1 were colored gray.

Expression of top differentially expressed genes in immune cell subtypes.

Differentially enriched immune cell types between responders and non-responders.

Analysis of ligand and receptor expression of immune modulators. Inferred ligand and receptor expression in the cancer and stroma compartments for immune checkpoint–related proteins (A) and cytokines (B). Heat maps are colored by the signed −log10(P-value) of differential expression between ICB responders and nonresponders (permutation test). Tiles with no differential expression (P-value >0.1) were colored gray. In the consensus heat maps, the median signed log10(P-value) were plotted for genes that were differentially expressed in three out of four ICB cohorts. The immune checkpoint pairs were annotated by the likely immune cell types as well as the effect of the checkpoint interaction (activating or inhibitory).

Differentially enriched immune cell types between responders and non-responders.

Identification of genes that are differentially expressed between responders and non-responders consistently across discovery cohorts.

Predictive modeling of response to immune checkpoint inhibition. A, Schematic of data and methods used in model training and testing. B, Top ranked features of ICB response selected by LASSO logistic regression. C, Comparisons of the performance of the IOselect model and the 3-feature model (TMB+IFNG+FCER1A) with existing biomarkers in six independent test cohorts. The differences between the mean AUCs of IOSelect and other models were computed using a paired t test. **, P-value <0.01; *, P-value <0.05; P-value <0.1. D, Kaplan–Meier curves of PFS for patients with high versus low predicted scores (greater than or less than median); P-values from log-rank test shown.

Validation of bulk tumor expression deconvolution in ICB-treated cohorts.