Diffuse large B-cell lymphoma (DLBCL) is a heterogeneous entity of B-cell lymphoma. Cell-of-origin (COO) classification of DLBCL is required in routine practice by the World Health Organization classification for biological and therapeutic insights. Genetic subtypes uncovered recently are based on distinct genetic alterations in DLBCL, which are different from the COO subtypes defined by gene expression signatures of normal B cells retained in DLBCL. We hypothesize that classifiers incorporating both genome-wide gene-expression and pathogenetic variables can improve the therapeutic significance of DLBCL classification. To develop such refined classifiers, we performed targeted RNA sequencing (RNA-Seq) with a commercially available next-generation sequencing (NGS) platform in a large cohort of 418 DLBCLs. Genetic and transcriptional data obtained by RNA-Seq in a single run were explored by state-of-the-art artificial intelligence (AI) to develop a NGS-COO classifier for COO assignment and NGS survival models for clinical outcome prediction. The NGS-COO model built through applying AI in the training set was robust, showing high concordance with COO classification by either Affymetrix GeneChip microarray or the NanoString Lymph2Cx assay in 2 validation sets. Although the NGS-COO model was not trained for clinical outcome, the activated B-cell–like compared with the germinal-center B-cell–like subtype had significantly poorer survival. The NGS survival models stratified 30% high-risk patients in the validation set with poor survival as in the training set. These results demonstrate that targeted RNA-Seq coupled with AI deep learning techniques provides reproducible, efficient, and affordable assays for clinical application. The clinical grade assays and NGS models integrating both genetic and transcriptional factors developed in this study may eventually support precision medicine in DLBCL.

Current use of liquid biopsy is based on cell-free DNA (cfDNA) and the evaluation of mutations or methylation pattern. However, expressed RNA can capture mutations, changes in expression levels due to methylation, and provide information on cell of origin, growth, and proliferation status. We developed an approach to isolate cell-free total nucleic acid (cfDNA) and used targeted next generation sequencing to sequence cell-free RNA (cfRNA) and cfDNA as new approach in liquid biopsy. We demonstrate that cfRNA is overall more sensitive than cfDNA in detecting mutations. We show that cfRNA is reliable in detecting fusion genes and cfDNA is reliable in detecting chromosomal gains and losses. cfRNA levels of various solid tumor biomarkers were significantly higher (P < 0.0001) in samples from solid tumors as compared with normal control. Similarly, cfRNA lymphoid markers and cfRNA myeloid markers were all higher in lymphoid and myeloid neoplasms, respectively as compared with control (P < 0.0001). Using machine learning we demonstrate cfRNA was highly predictive of diagnosis (AUC >0.98) of solid tumors, B-cell lymphoid neoplasms, T-cell lymphoid neoplasms, and myeloid neoplasms. In evaluating the host immune system, cfRNA CD4:CD8B and CD3D:CD19 ratios in normal controls were as expected (median: 5.92 and 6.87, respectively) and were significantly lower in solid tumors (P < 0.0002). This data suggests that liquid biopsy combining analysis of cfRNA with cfDNA is practical and may provide helpful information in predicting genomic abnormalities, diagnosis of neoplasms and evaluating both the tumor biology and the host response.

Diagnosis and classification of tumors is increasingly dependent on biomarkers. RNA expression profiling using next-generation sequencing provides reliable and reproducible information on the biology of cancer. This study investigated targeted transcriptome and artificial intelligence for differential diagnosis of hematologic and solid tumors. RNA samples from hematologic neoplasms (N = 2606), solid tumors (N = 2038), normal bone marrow (N = 782), and lymph node control (N = 24) were sequenced using next-generation sequencing using a targeted 1408-gene panel. Twenty subtypes of hematologic neoplasms and 24 subtypes of solid tumors were identified. Machine learning was used for diagnosis between two classes. Geometric mean naïve Bayesian classifier was used for differential diagnosis across 45 diagnostic entities with assigned rankings. Machine learning showed high accuracy in distinguishing between two diagnoses, with area under the curve varying between 1 and 0.841. Geometric mean naïve Bayesian algorithm was trained using 3045 samples and tested on 1415 samples, and showed correct first-choice diagnosis in 100%, 88%, 85%, 82%, 88%, 72%, and 72% of acute lymphoblastic leukemia, acute myeloid leukemia, diffuse large B-cell lymphoma, colorectal cancer, lung cancer, chronic lymphocytic leukemia, and follicular lymphoma cases, respectively. The data indicate that targeted transcriptome combined with artificial intelligence are highly useful for diagnosis and classification of various cancers. Mutation profiles and clinical information can improve these algorithms and minimize errors in diagnoses.

Multiple studies have demonstrated that diffuse large B-cell lymphoma (DLBCL) can be divided into subgroups based on their biology; however, these biological subgroups overlap clinically. Using machine learning, we developed an approach to stratify patients with DLBCL into four subgroups based on survival characteristics. This approach uses data from the targeted transcriptome to predict these survival subgroups. Using the expression levels of 180 genes, our model reliably predicted the four survival subgroups and was validated using independent groups of patients. Multivariate analysis showed that this patient stratification strategy encompasses various biological characteristics of DLBCL, and only TP53 mutations remained an independent prognostic biomarker. This novel approach for stratifying patients with DLBCL, based on the clinical outcome of rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone therapy, can be used to identify patients who may not respond well to these types of therapy, but would otherwise benefit from alternative therapy and clinical trials.

Most of the current utilization of liquid biopsy (LBx) is based on analyzing cell-free DNA(cfDNA). There is limited data on using cell-free RNA (cfRNA) levels (liquid transcriptome) in LBx. The major hurdles for using liquid transcriptome is its low level in circulation and the dilutional effects of various tissues that may pour their RNA into circulation. We explored the potential of using artificial intelligence (AI) to normalize the cancer-specific cfRNA and to enable liquid transcriptome to predict diagnosis. cfRNA transcriptomic data from 1009 peripheral blood samples was generated by hybrid capture next generation sequencing (NGS). Using two-thirds of samples for training and one third for testing, we demonstrate that AI is able to distinguish between normal control (N = 368) and patients with solid tumors (N = 404) with AUC = 0.820 (95 % CI: 0.760–0.879), patients with myeloid neoplasms (N = 99) with AUC = 0.858 (95 % CI: 0.793–0.924) and patients with lymphoid neoplasms (N = 128) with AUC = 0.788 (95 % CI: 0.687–0.888). Specific diagnosis was also possible when patients with lung, breast, colorectal, and myelodysplastic subgroups were tested. This data suggests that liquid transcriptomics when used with AI has the potential of transforming “liquid biopsy” to “true” biopsy, replacing the need for tissue biopsy.

Diffuse large B-cell lymphoma (DLBCL) is the most common type of lymphoma with high mutation burdens but a low response rate to immune checkpoint inhibitors. In this study, we performed targeted next-generation sequencing and fluorescent multiplex immunohistochemistry, and investigated the clinical significance and immunological effect of mutation numbers in 424 DLBCL patients treated with standard immunochemotherapy. We found that KMT2D and TP53 nonsynonymous mutations (MUT) were significantly associated with increased nonsynonymous mutation numbers, and that high mutation numbers (MUT^high) were associated with significantly poorer clinical outcome in germinal center B-cell-like DLBCL with wild-type TP53. To understand the underlying mechanisms, we identified a gene-expression profiling signature and the association of MUT^high with decreased T cells in DLBCL patients with wild-type TP53. On the other hand, in overall cohort, MUT^high was associated with lower PD-1 expression in T cells and PD-L1 expression in macrophages, suggesting a positive role of MUT^high in immune responses. Analysis in a whole-exome sequencing dataset of 304 patients deposited by Chapuy et al. validated the correlation of MUT-KMT2D with genomic complexity and the significantly poorer survival associated with higher numbers of genomic single nucleotide variants in activated B-cell–like DLBCL with wild-type TP53. Together, these results suggest that KMT2D inactivation or epigenetic dysregulation has a role in driving DLBCL genomic instability, and that genomic complexity has adverse impact on clinical outcome in DLBCL patients with wild-type TP53 treated with standard immunochemotherapy. The oncoimmune data in this study have important implications for biomarker and therapeutic studies in DLBCL.