Jeroen van Den Akker

Next generation sequencing (NGS) has become a common technology for clinical genetic tests. The quality of NGS calls varies widely and is influenced by features like reference sequence characteristics, read depth, and mapping accuracy. With recent advances in NGS technology and software tools, the majority of variants called using NGS alone are in fact accurate and reliable. However, a small subset of difficult-to-call variants that still do require orthogonal confirmation exist. For this reason, many clinical laboratories confirm NGS results using orthogonal technologies such as Sanger sequencing. Here, we report the development of a deterministic machine-learning-based model to differentiate between these two types of variant calls: those that do not require confirmation using an orthogonal technology (high confidence), and those that require additional quality testing (low confidence). This approach allows reliable NGS-based calling in a clinical setting by identifying the few important variant calls that require orthogonal confirmation. We developed and tested the model using a set of 7179 variants identified by a targeted NGS panel and re-tested by Sanger sequencing. The model incorporated several signals of sequence characteristics and call quality to determine if a variant was identified at high or low confidence. The model was tuned to eliminate false positives, defined as variants that were called by NGS but not confirmed by Sanger sequencing. The model achieved very high accuracy: 99.4% (95% confidence interval: +/− 0.03%). It categorized 92.2% (6622/7179) of the variants as high confidence, and 100% of these were confirmed to be present by Sanger sequencing. Among the variants that were categorized as low confidence, defined as NGS calls of low quality that are likely to be artifacts, 92.1% (513/557) were found to be not present by Sanger sequencing. This work shows that NGS data contains sufficient characteristics for a machine-learning-based model to differentiate low from high confidence variants. Additionally, it reveals the importance of incorporating site-specific features as well as variant call features in such a model.

Background: Biology has become the main driver of breast cancer therapy. Intrinsic biological subtypes by gene expression profiling have been identified. Pathology can be used to define surrogates of these subtypes but these are not always concordant, which may lead to different treatment plans. We investigated the concordance between BluePrint and MammaPrint (microarray-based) breast cancer subtypes and pathological surrogates (based on ER, PR, HER2 & Ki67). Methods: Using available data (centrally assessed pathology and genomics) from the MINDACT pilot phase [Rutgers et al. EJC 2011] 621 tumors were analyzed. Patients were classified according to 4-category based pathology (ER, PR, HER2 and Ki67); additionally, classification was performed adhering to the recent St. Gallen recommendations [Goldhirsch et al. 2011], which recognizes an additional category (Luminal B HER2+). Based on BluePrint and MammaPrint 4 subtypes are formed: Luminal A (Luminal-type/MammaPrint Low Risk); Luminal B (Luminal-type/MammaPrint High Risk); HER2-type; and Basal-type. This study is an analysis of discordant patient groups (i.e. clinical HER2+/BluePrint Luminal-type; clinical Hormone Receptor (HR)-positive/BluePrint Basal-type) providing comparison of centrally assessed tumor heterogeneity as well as comparison of quantified ER, PR and HER2 results. Results: Ki67 is often used as biomarker to distinguish Luminal A from Luminal B subgroups. The concordance between MammaPrint and centrally assessed Ki67 in Luminal-type patients is 71%, with a κ score of 0.35 (95% CI 0.26–0.45) indicating that Ki67 and MammaPrint cannot reliably substitute for each other. There is a relatively large group of clinical HER2+ cases that are BluePrint Luminal-type (29 out of 76; 38%) indicating that tumor expression of the Luminal profile is dominant compared with expression of the HER2 profile. These patients have high IHC ER results and fall into the group that St Gallen separately defines as Luminal B HER2-type. These patients may have lower response to trastuzumab [von Minckwitz et al. JCO 2012]. 12 out of 76 BluePrint Basal-type patients are clinical HR+. These patients have low centrally assessed IHC ER and PR expression (≥1% and Conclusions: Marked differences are observed between BluePrint and MammaPrint (microarray based) breast cancer subtypes and centrally re-assessed pathological surrogates (based on ER, PR, HER2 & Ki67). The greatest discordance is seen in the sub-stratification of Luminal patients, and in the HR+/HER2+ patients. The observed subtype discrepancies may have an important impact on treatment decision making. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P3-05-02.