Abstract In addition to structural chromosome aberrations resulting in fusion genes, the genome profile of haematological malignancies is characterized by copy number changes (gains and losses) relative to the ploidy level. Array comparative genome hybridization (aCGH)1 Combined with FISH and/or multiplex PCR array screening offers a reliable tool for assessment of genome complexity in haematological malignancies in diagnostic settings. The aCGH assay in our diagnostic algorithm uses 8 × 60K SurePrint G3 (Agilent, USA). The slides are scanned to measure the fluorescence ratio (FR), software evaluates imbalances. We recognize that the raw data in the form of log ratios, some 64,000 data points, contain sufficient information to identify the disease category without resorting to converting the log ratios to gains and losses and mapping them to individual chromosomes, a process open to operator interpretation. As an alternative and bearing in mind the current interest in machine learning, we attempted to auto-classify the data for the modest collection of 2,300 patient samples in our archive, according to their WHO classification. We were able to collect 175 samples for each group classified as myelodisplasia (MDS), acute myeloid leukemia (AML), myeloma (MM) and chronic lymphocytic leukemia (CLL), We used a Logistic Regression Classifier implemented in Mathematica 12. The classifier took only 2 minutes on average to complete the training. We took a total of 700 array sample files, each of 59,032 data points, of which 100 were randomly selected as a validation set. The balance of 600 files, distributed evenly and randomly over the 4 test diseases were used as a training set. We used a Confusion Matrix to evaluate and display the results as shown in the table. Confusion MatrixAMLCLLMDSMMTotalAML1609025CLL0150116MDS0116017MM1021417Predicted17162715 In spite of the small training set we correctly classified the MDS (16/17), CLL (15/16) and MM (14/17), 9 AML were reported as MDS a due to sharing genome imbalances. There was also confusion between the B cell disorders, MM and CLL known to carry overlapping genome aberrations. The auto classification offered by this method is a promising attempt to support a speedy and reliable assessment of genome complexity. There is no doubt that a better classifier would have been created with a larger training set. 1 Kallioniemi A et al., Kallioniemi OP, Sudar DA, Rutovitz D, Gray JW, Waldman F, Pinkel D (1992). “Comparative genomic hybridization for molecular cytogenetic analysis of solid tumors”. Science. 258 (5083): 818-821). Citation Format: Colin Grace, Temenuzhka Boneva, Elisabeth P. Nacheva. Successful classification of hematological neoplasm by array profile [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3641.