DEVELOPMENT OF CLINICALLY APPLICABLE NEXT-GENERATION BIOINFORMATICS: Driven by the Genomic Heterogeneity of Mantle Cell Lymphoma for Evidence-based Diagnostics, Disease Tracking, and Follow-up

MCL is a rare and malignant B-cell disease. It is often aggressive if not treated, but indolent cases do also occur. Although recent trials have shown long-term and ongoing remission following CAR T cell therapy for relapsing patients, MCL is still generally considered incurable. It is commonly identified by the characteristic expression of SOX11 or inter-chromosomal translocation between an immunoglobulin heavy chain gene enhancer/promoter and Cyclin D1, leading to a constitutive and elevated expression of the cell cycle regulator protein and thus enhancing B cell proliferation. While several recurrent mutations contribute to the oncogenesis of MCL, such as ATM mutations, lesions involving TP53 or the NOTCH1 pathway, the heterogeneity and small overlap between cases are striking and calls for a specialized strategy, aiming at systematic and high-resolution patient profiling. Furthermore, because MCL is defined by both a recurrent and heterogeneous spectrum of genomic lesions related to its oncogenic profile, it serves as a general model for improved profiling of individual cases in other B cell malignancies.

With MCL as a case model, this project aimed to investigate quality measures and signal distribution from whole-exome sequencing (WES) to leverage its clinical applicability. Through relatively simple concepts, the NGS workflow and bioinformatics are centered on practical molecular profiling of a few or individual patients, with the aim to support diagnosis and prognosis by maximizing the informational output.

The results derived from the studies showed that the number of false-positive coding somatic mutations in diagnosis-relapsed paired MCL efficiently decreased an order of magnitude by intersecting mutations present in its expressed counterpart, the mRNA, compared to the DNA alone. Thus, together with comprehensive annotation from clinically relevant databases, the cell purification and intersection of DNA and RNA provided a powerful tool for eliminating false-positive mutations.

While the sequencing reads of WES are not randomly distributed across the coding genome, the heterozygous variant allele frequencies of high-quality samples were approximately Gaussian. Therefore, it is found that the standard deviation provides an effective quality measure and can be employed in estimating the low burden of allelic imbalances. Furthermore, the developed methods for detecting copy-number alterations (CNA) show that the genomic juxtaposition of variants allele frequencies with relative gene-wise coverages provides efficient means for resolving acquired copy-neutral loss-of-heterozygosities. In other cases of MCL, with relapse in the central nervous system, I demonstrate that while sequencing batch effects are detrimental for detecting CNA, alternative techniques may circumvent such issues.

Through mathematical derivations, I present that detecting minute measurable disease by clonal rearrangements of immunoglobulin genes or somatic mutations is partly a stochastic problem. Thus, this approach helps explain why a potentially substantial amount of highquality DNA is required. Essentially, the amount of DNA or sequencing reads required for the confident detection of residual disease must generally exceed the aimed sensitivity level by several factors.

In conclusion, capture-based targeted sequencing of the coding genome is a powerful and consolidating tool for direct clinical implementation if applied critically. The inherent noise of WES demands high-quality samples and a high depth of coverage to be truly useful. Even a very high sequencing depth of targeted sequences will contain technical bias and stochastic fluctuations.