7075 Background: APOBEC (apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like) mutational signatures are uncommon in hematologic malignancies compared with solid tumors. This signature is characterized by C>T and C>G substitutions and clustered chromosomal abnormalities and is typically defined using whole-genome or whole-exome sequencing. Data on APOBEC signatures detected by targeted gene panels are limited. We interrogated a database of diffuse large B-cell lymphoma (DLBCL) cases sequenced between 2023 and 2025 using a targeted panel. Cases with >30 mutations were classified as “APOBEC-like,” and their molecular and expression features were analyzed. Methods: DNA and RNA were extracted from FFPE samples and sequenced by next-generation sequencing. DNA sequencing used a 302-gene panel and RNA sequencing a 1,600-gene panel. Hybrid-capture libraries were sequenced on an Illumina NovaSeq 6000. Results: Among 670 DLBCL cases, 39 (6%) demonstrated an APOBEC-like signature. A striking feature was the presence of multiple mutations within the same gene. One case harbored 20 distinct SOCS1 mutations, and 20 of 39 cases (51%) had at least one gene with ≥10 mutations. Cell of origin was germinal center B-cell in 21 cases (54%). All APOBEC-like cases showed numerous chromosomal abnormalities, most commonly 6q deletion/monosomy 6 (23 cases, 59%). Deletion of 17p was rare (2 cases). TP53 mutations were detected in 12 cases (31%), PIM1 in 23 (58%), and SOCS1 in 18 (46%). Gene expression profiling revealed significant differences between APOBEC-like and average DLBCL cases. TRAF3, TRAF5, and TRAF2 expression was significantly lower in APOBEC-like cases (log10 FDR < −12), while NAMPT, PRPF8, SF3A1, NFYC, LUC7L2-MCM3AP, and SMAD5 were significantly overexpressed (log10 FDR < −7). Despite the limited cohort size, random forest modeling demonstrated that expression profiling could distinguish APOBEC-like cases, with 20 genes achieving an AUC of 0.904 in a testing set. Conclusions: APOBEC-like mutational signatures can be identified in approximately 6% of DLBCL cases using targeted gene panels and are characterized by extreme intragenic hypermutation, recurrent chromosomal abnormalities, and a distinct expression profile that enables accurate classification by machine-learning approaches.
Albitar et al. (Wed,) studied this question.