A new scientific paper by a team led by Dr. Konrad Stawiski and Dr. Filipe L. F. De Carvalho has been published in the prestigious journal Briefings in Bioinformatics (Oxford University Press). The article presents the Somatic Likelihood Tiering (SLT) protocol – an innovative, fully interpretable, open-source bioinformatics tool that addresses the bottleneck in somatic variant analysis in tumor-only whole-exome sequencing (tumor-only WES). The work was conducted as part of an international collaboration between the Medical University of Łódź, Dana-Farber Cancer Institute, Brigham and Women’s Hospital, Harvard Medical School, and the Hospital del Mar Medical Research Institute in Barcelona.

Tumor-only whole-exome sequencing without matched normal tissue (e.g., a patient’s blood sample) is widely used in translational oncology, particularly for analyses of archival formalin-fixed, paraffin-embedded (FFPE) material, retrospective studies, and situations where access to reference samples is limited.

However, the lack of normal tissue creates a major analytical challenge. Standard variant-calling algorithms, such as Mutect2, can generate thousands of variants for a single patient. The overwhelming majority are unrelated to the tumor – they may be rare patient-specific germline polymorphisms, chemical fixation artifacts, or mutations resulting from clonal hematopoiesis of indeterminate potential (CHIP). Such a large number of variants makes efficient and precise clinical assessment by multidisciplinary tumor boards difficult.

Rather than relying on opaque, black-box machine-learning models, the SLT protocol organizes variants into four auditable prioritization tiers (SLT-A to SLT-D). Its performance was validated using the official international SEQC2 HCC1395 benchmark and two independent clinical cohorts (bladder cancer HdM-BLCA-1 and lung adenocarcinoma LUAD-TW). In the SEQC2 benchmark, the highest-priority tier (SLT-A) produced a concise list of 101 variants, including as many as 78 confirmed somatic mutations, achieving a positive predictive value (PPV) of 77.2%. The Number Needed to Review (NNR) for SLT-A was only 1.29, meaning that nearly four out of five reviewed variants were genuine somatic mutations. The study also demonstrated that the default minimum base-quality parameter in PureCN (BQ=25) silently discarded approximately 90% of variants in FFPE tissue. Applying the optimized threshold of BQ=20 restored full sensitivity of the cascade in archival material.

The research was initiated and carried out during Dr. Konrad Stawiski’s research fellowship at Dana-Farber Cancer Institute and Brigham and Women’s Hospital (Harvard Medical School) in Boston. The work received funding from the Polish National Agency for Academic Exchange under the Walczak Programme (grant no. BPN/WAL/2024/1/00014), the Medical Research Agency (grant no. 2023/ABM/02/00009), the National Cancer Institute (NCI), and the Bladder Cancer Advocacy Network. Key large-scale computational analyses were performed using the infrastructure of the Center for Digital Medicine at the Medical University of Łódź (MULDiMediC).

The SLT tool is fully open source (MIT License). Full publication (Open Access): doi:10.1093/bib/bbag487