J Clin Lab Anal

Background: Activating mutations in NOTCH1 are frequent in T-cell acute lymphoblastic leukemia (T-ALL) and, in the absence of alterations in RAS or PTEN, are associated with favorable prognosis. Besides classical heterodimerization and PEST domain mutations, juxtamembrane internal tandem duplications (JME-ITDs) represent a third class of activating variants. Their intermediate size and sequence composition can challenge short-read next-generation sequencing and variant calling algorithms, leading to underdetection in routine diagnostics.

Methods: An index case of newly diagnosed T-ALL harboring a NOTCH1 juxtamembrane insertion that was discordantly detected by variant callers during routine targeted NGS (Hematology OncoKitDx v2) prompted further analysis. The variant was evaluated using VarDict and VarScan2, followed by manual inspection with IGV and orthogonal confirmation by Sanger sequencing. To further assess variability in the detection of JME-ITDs, six additional BAM files from patients with previously confirmed variants were reanalyzed using VarDict, VarScan2, Mutect2, and Pindel.

Results: In the index case, VarDict detected a 51-bp NOTCH1 JME-ITD (c.5168-2_5216dup) that was missed by VarScan2 and confirmed by Sanger sequencing. Reanalysis of the seven cases showed that Mutect2 and VarDict detected all JME-ITDs (100%), VarScan2 identified only three (42.8%), and Pindel detected six mutations (85.7%). These results reflect differences in algorithmic strategies where haplotype-based callers incorporate soft-clipped and split reads, whereas pileup-based discard them.

Conclusion: Detection of NOTCH1 JME-ITDs depends strongly on variant-calling strategy. Combining haplotype-based callers with split-read structural variant tools may reduce false negatives and improve detection of clinically relevant insertions in diagnostic NGS pipelines.