Diagnosis of chronic B-cell lymphoproliferative disease in peripheral blood = how machine learning may help to the interpretation of flow cytometry data.

Flow cytometry (FCM) has become a method of choice for immunologic characterization of chronic lymphoproliferative disease (CLPD). To reduce the potential subjectivities of FCM data interpretation, we developed a machine learning random forest algorithm (RF) allowing unsupervised analysis. This assay relies on 16 parameters obtained from our FCM screening panel, routinely used in the exploration of peripheral blood (PB) samples (mean fluorescence intensity values (MFI) of CD19, CD45, CD5, CD20, CD200, CD23, HLA-DR, CD10 in CD19-gated B cells, ratio of kappa/Lambda, and different ratios of MFI B-cells/T-cells [CD20, CD200, CD23]). The RF algorithm was trained and validated on a large cohort of more than 300 annotated different CLPD cases (chronic B-cell leukemia, mantle cell lymphoma, marginal zone lymphoma, follicular lymphoma, splenic red pulp lymphoma, hairy cell leukemia) and non-tumoral selected from PB samples. The RF algorithm was able to differentiate tumoral from non-tumoral B-cells in all cases and to propose a correct CLPD classification in more than 90% of cases. In conclusion the RF algorithm could be proposed as an interesting help to FCM data interpretation allowing a first B-cells CLPD diagnostic hypothesis and/or to guide the management of complementary analysis (additional immunologic markers and genetic).

Adverse outcome in follicular lymphoma is associated with MYC rearrangements but not MYC extra copies.

Follicular lymphomas (FLs) with MYC rearrangements (MYC-R) and extra copies of MYC (MYC-EC) are rare and the prognosis impact is uncertain. We conducted a retrospective study including 321 FL patients, among whom 259 (81%) had no 8q24 alterations and 62 (19%) were assigned to 8qAlt. Forty-five cases were classified as MYC-EC and six as MYC-R. MYC-R patients were significantly older (P = 0·008), had higher follicular lymphoma international prognostic index (FLIPI) stage (P = 0·05) and ß2-microglobulin (ß2m; P = 0·05). Among patients treated with immuno-chemotherapy, four presented a MYC-R and 25 a MYC-EC. Univariate analysis showed the absence of significant difference between MYC-EC and normal MYC (MYC-NL) regarding progression-free survival (PFS; HR1·3; 95% CI [0·4-1·6]) and specific overall survival (SOS; HR 1·6; 95% CI [0·4-5·7]). Those results were compared to data from the PRIMA trial. This confirmed that MYC-EC had no impact on PFS (P = 0·86) or SOS (P = 0·9). Conversely, MYC-R was associated with a trend to inferior outcome regarding PFS (HR : 6·1; 95% CI [2·2-17·1]; P = 0·00026), lymphoma-related death (SOS; HR 13·6; 95% CI [2·9-65]; P = 0·00014) and risk of transformation (transformation-free survival (TFS); HR 82·7; 95% CI [14·8-463·4]; P < 0·0001). In conclusion, MYC-EC has no prognostic impact in FL but MYC-R FL tended to be associated with an increased risk of transformation and poorer outcome.

Clinical and biological features of B-cell neoplasms with CDK6 translocations: an association with a subgroup of splenic marginal zone lymphomas displaying frequent CD5 expression, prolymphocytic cells, and TP53 abnormalities.

A translocation involving the cyclin-dependent kinase 6 (CDK6) gene [t(CDK6)] is a rare but recurrent abnormality in B-cell neoplasms. To further characterise this aberration, we studied 57 cases; the largest series reported to date. Fluorescence in situ hybridisation analysis confirmed the involvement of CDK6 in all cases, including t(2;7)(p11;q21) immunoglobulin kappa locus (IGK)/CDK6 (n = 51), t(7;14)(q21;q32) CDK6/immunoglobulin heavy locus (IGH) (n = 2) and the previously undescribed t(7;14)(q21;q11) CDK6/T-cell receptor alpha locus (TRA)/T-cell receptor delta locus (TRD) (n = 4). In total, 10 patients were diagnosed with chronic lymphocytic leukaemia, monoclonal B-cell lymphocytosis or small lymphocytic lymphoma, and 47 had small B-cell lymphoma (SmBL) including 36 cases of marginal zone lymphoma (MZL; 34 splenic MZLs, one nodal MZL and one bronchus-associated lymphoid tissue lymphoma). In all, 18 of the 26 cytologically reviewed cases of MZL (69%) had an atypical aspect with prolymphocytic cells. Among the 47 patients with MZL/SmBL, CD5 expression was found in 26 (55%) and the tumour protein p53 (TP53) deletion in 22 (47%). The TP53 gene was mutated in 10/30 (33%); the 7q deletion was detected in only one case, and no Notch receptor 2 (NOTCH2) mutations were found. Immunoglobulin heavy-chain variable-region (IGHV) locus sequencing revealed that none harboured an IGHV1-02*04 gene. Overall survival was 82% at 10 years and not influenced by TP53 aberration. Our present findings suggest that most t(CDK6)+ neoplasms correspond to a particular subgroup of indolent marginal zone B-cell lymphomas with distinctive features.

Monoclonal B-cell lymphocytosis with a non-CLL immunophenotype - Review of 34 cases.

The monoclonal B-cell lymphocytosis (MBL) introduced as new entities in the 2008 WHO classification, are defined by circulating B-cell clone < 5.109/L without organomegaly and previous and/or simultaneous lymphoproliferative disorders. The MBL were subclassified in MBL CLL type (the most frequent), MBL atypical CLL type and MBL non-CLL type (rarely reported in literature). Here the clinic, cytologic, immunologic and genetic features of MBL non-CLL type were described from a series of 34 cases. As previously reported, present cases presented immunologic and genetic similarities to MZL and could be associated to the new proposed entity CBL-MZ (clonal B-cell lymphocytosis of marginal zone origin). In addition, few cases presented similarities to splenic diffuse red pulp lymphoma (SDRPL). In conclusion, according to the literature, MBL with non-CLL type (assimilated to CBL-MZ) may be a premalignant state of MZL and/or SDRPL.

del(8p) and TNFRSF10B loss are associated with a poor prognosis and resistance to fludarabine in chronic lymphocytic leukemia.

Chronic lymphocytic leukemia (CLL) is a heterogeneous disease, the prognosis of which varies according to the cytogenetic group. We characterized a rare chromosomal abnormality (del(8p), deletion of the short arm of chromosome 8) in the context of CLL. By comparing the largest cohort of del(8p) CLL to date (n = 57) with a non-del(8p) cohort (n = 155), del(8p) was significantly associated with a poor prognosis, a shorter time to first treatment, worse overall survival (OS), and a higher risk of Richter transformation. For patients treated with fludarabine-based regimens, the next-treatment-free survival and the OS were shorter in del(8p) cases (including those with mutated IGHV). One copy of the TNFRSF10B gene (coding a pro-apoptotic receptor activated by TRAIL) was lost in 91% of del(8p) CLL. TNFRSF10B was haploinsufficient in del(8p) CLL, and was involved in the modulation of fludarabine-induced cell death - as confirmed by our experiments in primary cells and in CRISPR-edited TNFRSF10B knock-out CLL cell lines. Lastly, del(8p) abrogated the synergy between fludarabine and TRAIL-induced apoptosis. Our results highlight del(8p)'s value as a prognostic marker and suggest that fit CLL patients (i.e. with mutated IGHV and no TP53 disruption) should be screened for del(8p) before the initiation of fludarabine-based treatment.

Deep Multi-Instance Learning Using Multi-Modal Data for Diagnosis of Lymphocytosis.

We investigate the use of recent advances in deep learning and propose an end-to-end trainable multi-instance convolutional neural network within a mixture-of-experts formulation that combines information from two types of data-images and clinical attributes-for the diagnosis of lymphocytosis. The convolutional network learns to extract meaningful features from images of blood cells using an embedding level approach and aggregates them. Moreover, the mixture-of-experts model combines information from these images as well as clinical attributes to form an end-to-end trainable pipeline for diagnosis of lymphocytosis. Our results demonstrate that even the convolutional network by itself is able to discover meaningful associations between the images and the diagnosis, indicating the presence of important unexploited information in the images. The mixture-of-experts formulation is shown to be more robust while maintaining performance via. a repeatability study to assess the effect of variability in data acquisition on the predictions. The proposed methods are compared with different methods from literature based both on conventional handcrafted features and machine learning, and on recent deep learning models based on attention mechanisms. Our method reports a balanced accuracy of [Formula: see text] and outperfroms the handcrafted feature-based and attention-based approaches as well that of biologists which scored [Formula: see text], [Formula: see text] and [Formula: see text] respectively. These results give insights on the potentials of the applicability of the proposed method in clinical practice. Our code and datasets can be found at https://github.com/msahasrabudhe/lymphoMIL.

Performances of Targeted RNA Sequencing for the Analysis of Fusion Transcripts, Gene Mutation, and Expression in Hematological Malignancies.

RNA sequencing holds great promise to improve the diagnostic of hematological malignancies, because this technique enables to detect fusion transcripts, to look for somatic mutations in oncogenes, and to capture transcriptomic signatures of nosological entities. However, the analytical performances of targeted RNA sequencing have not been extensively described in diagnostic samples. Using a targeted panel of 1385 cancer-related genes in a series of 100 diagnosis samples and 8 controls, we detected all the already known fusion transcripts and also discovered unknown and/or unsuspected fusion transcripts in 12 samples. Regarding the analysis of transcriptomic profiles, we show that targeted RNA sequencing is performant to discriminate acute lymphoblastic leukemia entities driven by different oncogenic translocations. Additionally, we show that 86% of the mutations identified at the DNA level are also detectable at the messenger RNA (mRNA) level, except for nonsense mutations that are subjected to mRNA decay. We conclude that targeted RNA sequencing might improve the diagnosis of hematological malignancies. Standardization of the preanalytical steps and further refinements of the panel design and of the bioinformatical pipelines will be an important step towards its use in standard diagnostic procedures.

Unclassifiable Isolated Monoclonal Lymphocytosis: Comprehensive Description of a Retrospective Cohort.

According to the World Health Organization (WHO) classification, the nosology of B-cell neoplasms integrates clinical, morphological, phenotypic, and genetic data. In this retrospective analysis, we identified 18 patients with isolated neoplastic lymphocytosis that could not be accurately classified within the WHO classification. Most of them were asymptomatic at the time of diagnosis and the evolution was relatively indolent, as only five patients required treatment after a median follow-up of 48 months. The neoplastic B-cells expressed CD5 in most cases, but the Royal Marsden Hospital score was strictly below 3. Trisomy 12 was the most frequent cytogenetic abnormality. High-throughput sequencing highlighted mutations found in both chronic lymphocytic leukemia (CLL) and marginal zone lymphoma (MZL). Similarly, the immunoglobulin heavy chain variable region repertoire was distinct from those reported in CLL or MZL. However, as treatment choice is dependent on the correct classification of the lymphoproliferative disorder, a histological diagnosis should be performed in case patients need to be treated.

Gain of the short arm of chromosome 2 (2p gain) has a significant role in drug-resistant chronic lymphocytic leukemia.

The different types of drug resistance encountered in chronic lymphocytic leukemia (CLL) cannot be fully accounted for by the 17p deletion (and/or TP53 mutation), a complex karyotype (CK), immunoglobulin heavy-chain variable region genes (IGHV) status and gene mutations. Hence, we sought to assess the associations between recurrent genomic abnormalities in CLL and the disease's development and outcome. To this end, we analyzed 64 samples from patients with CLL and gain of the short arm of chromosome 2 (2p+), which is frequent in late-stage and relapsed/refractory CLL. We found that fludarabine/cyclophosphamide/rituximab (a common first-line treatment in CLL) is not effective in removing the 2p+ clone - even in samples lacking a CK, the 17p deletion or unmutated IGHV. Our results suggest strongly that patients with CLL should be screened for 2p+ (using karyotyping and fluorescence in situ hybridization) before a treatment option is chosen. Longer follow-up is now required to evaluate bendamustine-rituximab, ibrutinib, and idelalisib-rituximab treatments.

Advances in the role of cytogenetic analysis in the molecular diagnosis of B-cell lymphomas.

INTRODUCTION: Cytogenetic abnormalities represent essential determinants of diagnosis and prognosis in B-cell lymphomas. Their theranostic value is increasingly significant with the development of targeted therapies, in order to adapt the treatment at diagnosis as well as when relapse occurs. Areas covered: As the significance of these biomarkers is influenced by the technology used to detect them, an overview describing the strength and weakness of conventional and emerging technologies is provided. This review also updates the diverse cytogenetic abnormalities found in B-cell lymphomas, emphasizing their value in treatment decision. Expert commentary: Cytogenetics remains an essential analysis for the diagnostic work-up of lymphomas. As whole genome sequencing becomes more and more affordable routinely, the next challenge will be to recover all the information conveyed by conventional karyotype, including the analysis of the clonal architecture at the single cell level, in whole genome data.