Genetic studies have shown that the common variant heritability of type 2 diabetes (T2D) risk and diabetes-related glycemic traits is enriched within the active regulatory regions of tissues important to diabetes and insulin resistance. Whole genome sequence (WGS) association analysis allows us to (1) fine-map known and novel loci across the non-coding genome utilizing these active regulatory regions, (2) determine the contribution of rare non-coding regulatory variants to the heritability of T2D risk and glycemic traits, and (3) perform rare variant tests of non-coding regulatory variants in addition to protein-coding variants. Here, we present a complete WGS analysis of fasting insulin (FI) and fasting glucose (FG) levels in phase 1 and 2 Trans-Omics for Precision Medicine (TOPMed) data (fasting glucose [FG] N=26,920 and fasting insulin [FI] N=23,361) with deep (>30x) sequence coverage in fourteen cohorts across five principle-component-derived ancestry groups. For our single variant analysis, we restricted analysis to variants with minor allele count > 20 in ancestry-specific and cross- ancestry pooled analysis. We used linear mixed effects models (R/Bioconductor GENESIS) and adjusted for sex, age and BMI, with empirical kinship for relatedness and population structure. For rare variant analysis, we derived aggregation units from coding variant annotations, regulatory annotations from pancreatic islets, liver, muscle and adipose tissues, and annotation Principal Components (aPCs), a multi-dimensional summary of functional annotation scores. We used sequence kernel association test (SKAT) and the novel variant-Set Test for Association using Annotation infoRmation (STAAR), a general framework that incorporates annotation scores using an omnibus multi-dimensional weighting scheme. In single variant analysis, we observe 8 regions associated (P<5x10-8) with FG and 6 regions associated with log-FI, with most signals from common variants in known regions. In rare variant analysis, one test showed a (Bonferroni) significant association with FG using missense variants in G6PC2. In G6PC2, multiple distinct associations were observed in the single variant tests, consistent with previous reports, and nominal significant associations were observed when aggregating predicted loss-of- function rare variants. Analysis with larger samples are planned. Our work refines the genetic architecture of glycemic traits using the full spectrum of genetic variation.