e19574 Background: Lymphoma is common cancer worldwide, a large group of lymphoid hematopoietic malignancies including Non-Hodgkin lymphoma (NHL) and Hodgkin lymphoma (HL) two major type. Lymphoma includes more than 30 unique subtypes, originating from B, T and NK cells. Currently, diagnosis and classification of lymphoma is based on excisional lymph node biopsy through immunohistochemistry and in situ hybridization. However, invasive biopsies have significant limitations and carry procedural risks to patients, and it cannot account for spatial inter- and intra-tumor heterogeneity due to sampling from only one location in a single tumor lesion. In the past few years, an emerging class of methods named Liquid biopsy can potentially improve on these limitations. Methods: 135 untreated stage I-IV lymphoma patients and 399 healthy individuals were enrolled in this study. 8ml peripheral blood was collected from each participant after enrollment. Cell-free DNA (cfDNA) was extracted and subjected to shallow whole-genome sequencing (sWGS), and partly plasma was used to quantify the expression of 7 tumor serum protein markers. We developed a newly blood-based assay, which integrated the newly next-generation technology of sWGS of cfDNA and protein marker (CA125), with artificial intelligence (AI) technology. Though our multi-dimensional assay, the cancer risk score (CRS) of each sample were calculated, and was used for lymphoma early diagnosis. Results: After compared each protein marker between lymphoma and healthy groups, only CA125 could be used to screen the lymphoma with a sensitivity of 25.9% at extremely high specificity 98.0% and an area under the curve (AUC) of 69.5%. As for the genomic feature of cfDNA, the characteristics of copy number aberrations (CNA) and fragment size (FS) were significantly different between lymphoma and healthy groups (p value < 0.0001 by student’s t-test). Both FS and CNA demonstrated improved sensitivity (36.3% and 67.4% respectively) compared to CA125. When the three signatures were incorporated into the CRS model, it achieved the best performance allowing 95 lymphoma cases to be identified with a sensitivity of 70.4% at 98.0% specificity, and an AUC of 88.0%. In stage I/II and III/IV lymphoma cases, the sensitivity was 40.0% and 85.7% respectively. For each lymphoma subtype, the lymphoma which from B-cell had higher sensitivity than that from NK/T cell (71.0% vs 64.3%), especially in HL (91.7%), all originated from B cell. Conclusions: We newly find the traditional protein marker CA125 can be used for lymphoma screening, which was integrated with genomic feature CNA and FS of cfDNA as a multidimensional assay for early detection of lymphoma with sufficient accuracy. The performance of our efficient and non-invasive assay for detecting lymphoma is 70.4% sensitivity at extremely high specificity 98.0%, especially for 40.0% sensitivity for early patients (stage I/II) screening.
Read full abstract