Machine learning-based analysis has been recently applied to image data acquired in imaging flow cytometry technologies including both analyzers and sorters. Compared to analysis of conventional microscopy images, the largest difference appears when we have to complete the analysis within a limited time scale for accomplishing realtime analysis and subsequent cell sorting. In this article, we propose to categorize the analysis approaches into two groups based on the type of data modality, raw imaging signals or features explicitly extracted from images, being analyzed by a trained model. We hope that this review helps readers understand what kind of uniqueness, differences, and opportunities shows up depending on the way of implementing machine learning-based analysis in the recently developed “imaging” cell sorters.

One-carbon metabolism, also called folate-mediated metabolism, has been targeted for the treatment of patients with cancer. However, therapeutic windows available to the inhibitors of one-carbon metabolism are limited and some patients have been shown to develop resistance to anti-folate drugs. Recently, we showed that mitochondrial enzymes involved in one-carbon metabolism have greater specifi city toward cancer cells than the cytoplasmic enzymes. In our study, we knocked down the methylenetetrahydrofolate dehydrogenase (NADP+ dependent) 2, methenyltetrahydrofolate cyclohydrolase (MTHFD2) gene in order to inhibit the activity of one such mitochondrial enzyme. Cell cycle progression was analyzed using propidium iodide, and stem cell-like nature was evaluated in terms of aldehyde dehydrogenase activity on a fl ow cytometric setup. Our results showed a drastic decrease in cellular proliferation and stem cell-like phenotypes in the MTHFD2 knockdown cells. From a mechanistic point of view, the inhibition of cellular growth could be predominantly ascribed to the depletion of purine nucleotides, and that of stem cell-like phenotypes was attributable to the accumulation of 1-(5'-phosphoribosyl)-5-amino-4- imidazolecarboxamide.

Since the mouse models with the driver mutations of myeloproliferative neoplasms (MPNs) reproduce various pathological conditions found in MPNs patients such as blood cell increase, splenomegaly, and bone marrow fi brosis, these mouse models are extremely useful as tools to elucidate the mechanism of pathogenesis. Activation of MPL and its downstream signaling pathways is common to the JAK2, MPL, and CALR mutations, which are major driver mutations in MPNs, and is essential for the pathogenesis of MPNs. We analyzed the gene expression profi le of MPLexpressing cells in Jak2V617F transgenic mice and identified the signaling pathway responsible for bone marrow fi brosis. MPL activates the transcription factor upstream transcription factor via the p42/44 MAPK pathway, and the JAK2V617F mutant protein uses this pathway to enhance transcription of TGF-β1, a cytokine essential for bone marrow fibrosis. Inhibition of this pathway effectively suppressed organ fibrosis in Jak2V617F mice in vivo. JAK2 inhibitors have only a slow and limited effect on improving myelofibrosis. The mechanism we have discovered is a promising therapeutic target for antifi brotic therapy, which has recently been developed for patients with myelofi brosis. Such anti-fi brotic therapies may create new therapeutic benefi ts for patients with myelofi brosis that cannot be provided by JAK2 inhibitors. Since the mouse models with the driver mutations of myeloproliferative neoplasms (MPNs) reproduce various pathological conditions found in MPNs patients such as blood cell increase, splenomegaly, and bone marrow fi brosis, these mouse models are extremely useful as tools to elucidate the mechanism of pathogenesis. Activation of MPL and its downstream signaling pathways is common to the JAK2, MPL, and CALR mutations, which are major driver mutations in MPNs, and is essential for the pathogenesis of MPNs. We analyzed the gene expression profi le of MPLexpressing cells in Jak2V617F transgenic mice and identified the signaling pathway responsible for bone marrow fi brosis. MPL activates the transcription factor upstream transcription factor via the p42/44 MAPK pathway, and the JAK2V617F mutant protein uses this pathway to enhance transcription of TGF-β1, a cytokine essential for bone marrow fibrosis. Inhibition of this pathway effectively suppressed organ fibrosis in Jak2V617F mice in vivo. JAK2 inhibitors have only a slow and limited effect on improving myelofibrosis. The mechanism we have discovered is a promising therapeutic target for antifi brotic therapy, which has recently been developed for patients with myelofi brosis.

Such anti-fi brotic therapies may create new therapeutic benefi ts for patients with myelofi brosis that cannot be provided by JAK2 inhibitors.

Flow cytometric immunophenotyping is an essential methodology for establishing the diagnosis and prognostic classifi cation and for detecting measurable residual disease (MRD) in acute myeloid leukemia (AML). Recently, more sophisticated approaches have been needed in fl ow cytometry (FCM), owing to the improved ability to assess between phenotypes and genetic abnormalities and detect small malignant populations based on aberrant antigen expression at diagnosis. We studied 149 patients with newly diagnosed de novo AML and evaluated the utility of multi-parametric FCM (MFCM) for diagnosing and detecting MRD. Our approach was capable of not only determining the lineages and differentiation status of neoplastic cells but predicting the presence or absence of genetic abnormalities. Moreover, by using a patient-specifi c antibody panel based on aberrant antigen expression, we could detect MRD with high sensitivity (up to 10^-4).