Abstract Background In complex automated clinical laboratory Next Generation Sequencing (NGS) workflows, numerous opportunities exist for contamination events to occur. For oncology testing applications, detection of these events is crucial for accurate results reporting. A contamination detection module (MICon) using microhaplotype (MH) regions and an in-house designed analysis model was developed for use in a NGS assay for myeloid neoplasms (NGSHM) [1]. Variant detection for NGSHM has a validated analytical sensitivity of 2% variant allele frequency (VAF). Gross contamination events can potentially cause erroneous false positive variant detection whereas true low VAF mutations may be masked and evade detection. The following study was performed to evaluate MICon module performance on samples processed over the course of a year. Methods MH regions contain highly conserved tandem single nucleotide polymorphisms (SNPs) with high global heterogeneity occurring within a 300-nucleotide span. From the work of Kidd et al [2], 27 MH regions covering 92 SNPs and spanning 16 chromosomes were selected for inclusion in the 47 gene NGSHM target-capture panel. Sequencing was performed on a NovaSeq 6000 and analyzed through an internal bioinformatics pipeline, which incorporates the MICon module. MICon uses a binary classification model incorporating VAFs from MH regions, number of MH genotypes, and the contamination estimation score from verifyBamID to compute a single value on a 0-100 scale. Scores <50 are classified as non-contaminated and scores ≥50 require further investigation. Results The dataset was comprised of 10,990 samples. A total of 10,232 patient cases (93.1%) had non-contaminated MICon scores <50. Control replicates accounted for 261 samples (2.4%). There were 497 patient cases (4.5%) with MICon scores ≥50. Of the 497 patient cases, 137 (27.6%) had undergone hematopoietic stem cell transplantation (HSCT) and 114 (22.9%) contained gross chromosomal abnormalities, representing iatrogenic or tumor biological factors. An additional 135 (27.2%) of cases were contaminated from laboratory processing errors, 45 (9.1%) occurred due to chemistry failures during run processing, and 66 (13.2%) had an undetermined cause for a high score, representing laboratory-related and unknown factors. Overall, the 135 laboratory processing errors detected accounted for 1.3% of patient cases analyzed by the laboratory. These cases were repeated, and the non-contaminated results were accurately reported. Conclusions Herein it was shown the implementation of MICon has improved patient safety over the course of a year. The 1.3% of patient cases identified as arising from laboratory processing errors were successfully reprocessed. While the overall rate of laboratory processing errors appears low, the absolute number can be significant in a high-volume test setting, despite reliance on highly automated processes. Value is added by MICon through identifying errors that could result in the release of inaccurate patient results. By detecting previously unrealized errors, MICon has become an invaluable asset to the NGSHM assay by enhancing patient laboratory testing safety.
Read full abstract