ALBERT

Description: Background Several recurrent somatic mutations have been identified in MDS and these mutations play an important role in disease pathophysiology and outcome. BCOR and BCORL1 are located on chromosome X and interact with histone deacetylases and other cell functions. The BCOR gene is mutated (BCORMUT) in 4-6% of MDS patients (pts) and is associated with poor outcome. BCORL1 mutations (BCORL1MUT ) are present in

Description: Background: While treatment with the hypomethylating agents (HMAs) azacitidine (AZA) and decitabine (DAC) improves cytopenias and prolongs survival in MDS patients (pts), only 30-40% of pts respond. Genomic and/or clinical models that can predict which pts will respond could prevent prolonged exposure to ineffective therapy, avoid toxicities and decrease unnecessary treatment costs. Machine learning (ML), a field of artificial intelligence, is an advanced computational analysis of complex data sets that can overcome some of the limitations of standard statistical methods. ML uses computational algorithms to automatically extract hidden information from a dataset by learning from relationships, patterns, and trends in the data. Thus, ML can produce powerful, reliable and reproducible predictive models based on large and complex datasets. The aim of this project is to build a geno-clinical model that uses ML algorthims to predict responses to HMAs. Methods: We screened a cohort of 433 pts with MDS who received HMAs at multiple academic institutions for the presence of common myeloid somatic mutations in 29 genes. Responses were assessed per International Working Group 2006 criteria. Five popular supervised classification ML algorithms including: random forest (RF), tree ensemble (TE), naive bayes (NB), decision tree (DT), and support vector machine (SVM) algorithms were used individually and in combination to enhance the accuracy of the proposed model (bag of model approach). For each iteration, the dataset was divided randomly into training and validation cohorts. The partition of the dataset was repeated multiple times randomly to minimize biases in pt selection. A 10-fold cross validation was also used on the entire dataset to assure data reproducibility. Important variables were selected using backward feature elimination and tree depth scores. Performance was evaluated according to the area under curve (AUC) and accuracy matrix. All analyses were done using KNIME (an open analytic platform for ML). Results: Among 433 pts, 193 (45%) received AZA, 176 (40%) DAC, and 64 (14%) received HMA +/- combination. The median age was 70 years (range, 31-100) and 28% were females. Responses included: 95 (58%) complete remission (CR), 14 (3%) marrow CR, 16 (4%) partial remission (PR), and 59 (14%) hematologic improvement (HI). For the purpose of this analysis, pts with CR/PR/HI were considered as responders. The most commonly mutated genes were: ASXL1 (31%), TET2 (22%), SRSF2 (17%), RUNX1 (15%), and DNMT3A (14%). In univariate analyses, no single mutation was more prevalent in responders compared to non-responders except NF1 (more common in non-responders, p = .04). A logistic regression multivariate analysis did not produce a reliable and reproducible model. When applying ML algorithms on learner (80% randomly selected pts) and predictor cohorts, the accuracy rate in predicting responses for RF was 64%, for TE 60%, for NB 60%, for DT 66%, and for SVM 51%. When results from each model were combined (a bag of models approach), the accuracy increased to 69%. Backward feature elimination and tree depth scores identified the following factors as predictors of response: hemoglobin 69 years, TP53 with variant allelic frequency (VAF) 〉15%, CBL VAF 〉30%, and RUNX1 VAF 〉 25%. Only ASXL1mutations at any VAF were predictive of HMA resistance. Interestingly, none of the mutations were selected for response or resistance when the models did not include VAF. Neither treatment modality with azacitidine vs. decitabine vs. combination nor treatment center impacted response. When the analysis was restricted to pts with higher-risk disease by IPSS, the accuracy rate in predicting responses improved: for RF it became 71%, for TE 65%, for NB 60%, for DT 64%, and for SVM 76%. When the analysis was focused on pts who achieved CR vs. No CR, the models predicted the response differently. The RF and TE models were able to predict No CR with an accuracy rate of 75% and 76% respectively. Other models were able to predict CR and No CR with lower accuracy. Conclusion: We propose a novel geno-clinical model that uses machine intelligence to predict HMA response/resistance in pts with MDS. The model has a higher accuracy rate in higher-risk MDS pts. ML can open opportunities in translating genomic data into reliable predictive models that can aid physicians in clinical decision making. Disclosures Bejar: Celgene: Consultancy, Honoraria; Foundation Medicine: Consultancy; Genoptix: Consultancy, Honoraria, Patents & Royalties: No royalties.

Description: Background: Sideroblastic anemia (SA) can present as congenital SA (CSA), acquired clonal SA, and acquired reversible SA. Patients (pts) with SA have anemia and ring sideroblasts (RS). Acquired clonal SA is often linked to myelodysplastic syndromes (MDS) or myelodysplastic/ myeloproliferative neoplasms (MDS/MPN). Clinico-pathologic overlap features, unmet morphologic and/or cytogenetic criteria complicate the diagnosis of SA leading to delayed therapies. Currently the diagnosis of SA is based on bone marrow (BM) examination and routine blood tests. There is a need to find easily testable biomarkers that can lead to faster diagnosis of clonal and non-clonal SA. Somatic mutation in splicing factor 3b, subunit 1 (SF3B1) are frequent in MDS-RS and MDS/MPN-RS and have been closely associated with RS. Objective: SF3B1 mutations can be a useful diagnostic biomarker for pts with acquired clonal SA who present with cytopenias and/or minimal morphologic changes suspicious of MDS and MDS/MPN but not sufficient to make a definitive diagnosis. Patients and Methods: Six pts with SA at presentation and seen at Cleveland Clinic were included in this study. The median age was 38 years (range, 6-75). Blood tests and BM biopsy showed persistent anemia [Hgb, 10.5 g/dL (range, 8.8-13)], RS [numerous (3 pts), 15% (1 pt), rare (1 pt) and 〉15% (2 pts)], 3/6 pts had minimal erythrodysplasia with 1 pt having a mild megakaryocytic dysplasia, 3 pts had hypercellular (60-90%), 2 pts had normocellular (50%, 80%) and 1 pt had hypocellular BM (30%) for age, and 〈 5% BM (2%=2 pts; 1%=1 pt; 3%=1 pt). Two pts had PLT count 〉 400 k/uL, 3 pts had 〉 100 k/uL, and 1pt

Description: Background Several prognostic models have been developed to risk stratify patients (pts) with MDS including: the International Prognostic Scoring System (IPSS), Revised IPSS (IPSS-R), World Health Organization classification-based Prognostic Scoring System (WPSS), and MD Anderson Prognostic Scoring System (MDPSS). All except for the MDPSS were developed in treatment-naive pts and, if validated in treated pts, were done so primarily in those receiving one line of therapy. Incorporation of molecular data into the IPSS-R improves its predictive power, and adding molecular data to the IPSS can upstage or downstage some pts. In this study, we compared the prognostic utility of each prognostic model, after adding molecular data, in treated MDS pts. Method Clinical and mutational data from MDS pts diagnosed between 1/2000-1/2013 were analyzed. A panel of 60 gene mutations that were described as commonly mutated in myeloid malignancies was included. Patients who underwent hematopoietic cell transplant (HCT) were censored at the time of transplant. Univariable and multivariable analyses on the training cohort were performed by applying Cox proportional hazards regression analyses that included age, model score, and molecular mutations with an outcome of overall survival (OS). All molecular models were then applied to the validation cohort. The fit of the proposed models to the data was assessed by using Akaike's information criterion (AIC, lower values imply better fit) and concordance (c-) index. Results A total of 610 pts were included and divided into two cohorts, training (404 pts), and validation (206 pts). Median age of the training cohort was 67 years (range, 19-88); 83 pts (20%) had MDS/MPN including chronic myelomonocytic leukemia (CMML). Pts received a median of 2 lines of therapies (range, 0-7) and 15% of pts underwent HCT. First line therapies included: supportive care (22%), growth factors (22%), azacitidine +/- combinations (30%), decitabine +/- combinations (6%), lenalidomide (5%), induction chemotherapy (3%), immunosuppressive therapy (3%), and other therapies /clinical trials (9%). The median OS in the training and validation cohorts per scoring system (SS), i.e. IPSS, WPSS, MDPSS, and IPSS-R, is summarized in Table 1. The most common mutations in the training cohort were: TET2 (19%), ASXL1 (16%), SF3B1 (13%), DNMT3A (9%), STAG2 (9%), RUNX1 (8%), and U2AF1 (7%). In univariate analyses, mutations in EZH2 (HR1.8, p = .02), TP53 (HR2.3, p 〈 .001), RUNX1 (HR 1.5, p = .05), and NPM1 (HR1.49, p = .001) had a negative impact on OS while SF3B1 (HR .34, p 〈 .001) mutations were associated with favorable outcome. In multivariate analyses that included SS, mutations (from the list above), and age (except for the MDPSS, which already includes age), the following independent prognostic factors for OS were identified: age, EZH2, SF3B1, TP53, and each SS. Based on the fitted coefficients of each prognostic factor, a molecular version of each model including IPSSm, WPSSm, MDPSSm, and IPSS-Rm was proposed with median OS in the training and validation cohorts as summarized in Table 1. The addition of molecular data improved (reduced) the AIC and (raised) the C-index for the IPSS (2332.6, .69 vs. 2365.8, .64), WPSS (2322.3, .68 vs. 2356.2, .65), MDPSS (2308.4, .71 vs. 2323.4, .69), and IPSS-R (2305.2, .70 vs. 2330.5, .66), respectively. Further, the addition of molecular data to the IPSS upstaged 37% of pts from lower- to higher-risk disease and downstaged 5% of intermediate-1 to low risk disease. In the WPSS, it upstaged 21% of pts and downstaged 24%; in the MDPSS it upstaged 19% and dowstaged 22% of pts from intermediate-1 to low risk; and in the IPSS-R it upstaged 26% to higher-risk disease and 59% of pts with intermediate risk to a higher risk category. Conclusions The addition of molecular data to established MDS prognostic models can improve their predictive power, even in treated MDS patients. More importantly, adding molecular information can upstage or downstage pts into different risk categories. This study highlights the importance of incorporating molecular data into clinical prognostic systems. Disclosures Mukherjee: Celgene: Consultancy, Honoraria, Research Funding; Novartis: Consultancy, Honoraria, Research Funding; Ariad: Consultancy, Honoraria, Research Funding.

Description: The identification of the JAK2V617F mutation in myeloproliferative neoplasms (MPN) paved the way for the pivotal studies that led to the FDA approval of a JAK1/2 inhibitor, ruxolitinib (rux) in patients (pts) with myelofibrosis (MF). Improvement in splenomegaly and debilitating disease-related symptoms were the primary clinical responses observed with rux. Although JAK2 mutational status did not impact response/survival in MF pts, cytogenetics had an impact on prognosis. In a related myeloid neoplasm specifically myelodysplastic syndromes, molecular mutations (TET2/DNMT3A) predict for better therapeutic response to DNA methyltransferase inhibitors. We hypothesized that somatic mutations and single nucleotide polymorphism array (SNP-A) lesions are frequent in MF pts treated with rux and may affect their clinical outcomes. To further investigate the predictive and prognostic impact of SNP-A lesions and somatic mutations in MF pts in the rux era, we studied 54 MF pts who received at least 12 weeks of rux therapy (tx) using a modified dose escalation approach (Tabarroki A et al. 55th ASH; Abstract 1586). Clinical (total symptom score [TSS], spleen size), cytogenetic (metaphase cytogenetics [MC], SNP-A), hematologic and survival data were collected before and 12-weeks post rux tx. Categorical data were analyzed using X2 test. A p-value of

Description: Objectives: TP53, a tumor suppressor gene, is frequently mutated in myeloid malignancies. The negative impact of TP53 mutations in myelodysplastic syndromes (MDS) has been described previously but there is controversy regarding the prognostic impact of the mutation's characteristics (location, type, passenger vs. driver, and others). Methods: We sequenced DNA samples from 732 patients (pts) with MDS and related myeloid malignancies for the presence of TP53 mutations and 61 other genes that have been described as recurrently mutated in MDS. Overall survival (OS) was measured from the time of diagnosis to time of death or last follows up. Variant allele frequencies (VAFs) adjusted by zygosity were used to define clonal architecture of driver clones. Results: Of 80 mutations detected in 73 (10%) pts, 66 (88%) were missense, 7 (9%) were nonsense, and 7 (9%) were frame shift deletions/insertions. Pts with TP53 mutations had a higher WBC (4.6 vs. 3.9 X 109/L, p = .04), higher bone marrow blast % (median 9 vs. 3, p =0.1), and a higher risk category by revised International prognostic Scoring System (56% vs. 27%, p =.01) compared to pts with TP53 wild type. TP53 mutations were commonly occurred with TET2 (16%), PRPF8 (13%), ASXL1 (11%), DDX54 (8%), DNMT3A (8%), and IDH2 (8%). The mean VAF for TP53 was 41.9 (5-100). TP53 mutations were defined as drivers in 20% of samples, passengers in 40%, and mosaic in 40%. Mutation positions included: 19 (24%) in the DNA binding domain, 2 (3%) in the transactivation domain, 1 (1%) in the tetramerization domain and 58 (72%) other. With a median follow up of 16.4 months, the median OS for the entire group was 8.24 months. Patients with TP53 as driver mutations had a worse OS compared with patients with TP53 as passenger mutations (median, 2.2 vs. 13.0 months, respectively, p =.02). Similarly, OS by TP53 VAF categorized as low (50%) was 12.4, 8.5, and 3.4 months, respectively. Among patients with available treatment data, 29 patients were treated with the hypomethylating agents, azacitidine or decitabine, 17% responded (60% CR, and 40% PR), and OS was similar compared to patients treated with other therapies. Patients treated with Hematopoietic stem cell transplant (HCT) had superior OS compared to pts not receiving HCT (median, 14.9 vs. 8.9 months, respectively, p =.05). Conclusions: TP53 mutations are associated with poor outcome in MDS, but the outcome varies depending on the type of mutation and VAF. Treatment with HCT remains a valid treatment option in a subset of patients but novel treatment strategies are desperately needed. Disclosures Mukherjee: Novartis: Consultancy, Honoraria, Research Funding; Ariad: Consultancy, Honoraria, Research Funding; Celgene: Consultancy, Honoraria, Research Funding.