Clinical validation of the tempus xT next-generation targeted oncology sequencing assay

Abstract

We developed and clinically validated a hybrid capture next generation sequencing assay to detect somatic alterations and microsatellite instability in solid tumors and hematologic malignancies. This targeted oncology assay utilizes tumor-normal matched samples for highly accurate somatic alteration calling and whole transcriptome RNA sequencing for unbiased identification of gene fusion events. The assay was validated with a combination of clinical specimens and cell lines, and recorded a sensitivity of 99.1% for single nucleotide variants, 98.1% for indels, 99.9% for gene rearrangements, 98.4% for copy number variations, and 99.9% for microsatellite instability detection. This assay presents a wide array of data for clinical management and clinical trial enrollment while conserving limited tissue.

Keywords: next-generation sequencing assay validation; tumor profiling.

Figure 1. Performance of SNV and Indel Detection by the Tempus xT assay

(A) Precision of indel detection by VAF. Precision was calculated for each bin of variants with allele fractions rounded to the nearest 5 percent. The vertical black line corresponds to the LOD. (B) Precision of SNV detection by VAF, as in A. (C) Correlation of xT assay indel VAFs to xO assay indel VAFs. (D) Correlation of xT assay SNV VAFs to xO assay SNV VAFs. (E) VAFs of indels and SNVs detected on chromosome 17 in four samples with serial 1:1 dilutions of the xT assay. Dark blue lines indicate best fit of a linear model. (F) Positive control detection. Boxplots of the VAF of three SNVs in a positive control sample run on every xT assay over a period of 5 months. The single point marked in black is an AKT1 p.E17K variant which failed filtering criteria.

Figure 2. Analysis of rearrangement detection performance by the Tempus pipeline and retrospective analysis of recurrent fusions

(A) Fusions detected by cancer type. (B) Positive control fusion samples (ROS1-SLC34A2) processed by the xT assay over the course of 4 months. The fusion was expected at 5% VAF in the control and was consistently detected across flow cells and instruments. (C) Limit of detection analysis for two serially diluted fusions (ALK-EML4 and RET-ANKRD26). Both fusions were detected down to 3–5% simulated VAF by the Tempus xT Assay. (D) Functional characterization of fusions detected by the xT assay. Domains, regions, and sites are highlighted for orientation along the amino acid sequence for each protein involved in the rearrangement event. (E) Analysis of the recurrent TMPRSS2-ERG fusion found in 13 prostate cancers detected by the xT assay. The xT assay consistently localized breakpoints to the expected functional domains resulting in the TMPRSS2 promoter and replacing the first several exons of ERG by a chromosomal deletion. This results in the functional domains of ERG being largely intact, but under the control of the TMPRSS2 promoter.

Figure 3. Analysis of performance of the xT assay in the detection of copy number variations

(A) Detection comparison of Tempus copy number calls against validated ERBB2 control ratios. (B) Plot of reported CNVs by cancer type for the xT patient cohort sequenced at Tempus. (C) A representative 1p-19q co-deletion detected commonly in oligodendrogliomas. The blue dots represent genomic regions showing one copy loss while the grey points represent neutral segments.

Figure 4. Survey of immunotherapy markers across diverse cancer types

(A) The distribution of TMB for each cancer type plotted on a log2 scale and ordered by the median TMB. Outliers (data points beyond 1.5x interquartile range) are shown as individual points. (B) Analysis of samples in the top 10th percentile of TMB. The inset shows the distribution of TMB across all samples included in the study, with the vertical bar marking the 90th percentile of TMB. The bar chart shows the proportion of predicted antigenic mutations of the non-synonymous mutations detected. MSI-H samples are highlighted in red. The color-coded matrices show the MMR gene mutations detected by mutation type (top), predicted MLH1 methylation status (middle), and DNA repair gene mutations detected (bottom).

Figure 5. Mutational landscape across all cancer types

(A) Plot of a subset of patients within the Tempus xT cohort containing at least one biologically relevant alteration, stratified by alteration prevalence. Patients were clustered by the mutational profile similarity of genes (y-axis) with at least 5 actionable alterations detected within the cohort. Patient cancer type is displayed by the colored bar below the matrix. (B) Lollipop plot of TP53, KRAS, and PIK3CA alterations detected by the xT assay across all cancer types.