Prediction of myeloid malignant cells in Fanconi anemia using machine learning

Abstract

ABSTRACTFanconi anemia (FA) is an inherited bone marrow failure syndrome with cancer predisposition. Most FA patients develop aplastic anemia during childhood and have an extremely high cumulative risk to develop cancer during their lifespan. Myeloid malignancy is one of the main tumor risks for patients with FA, including high-risk myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML). Although bone marrow transplantation is the treatment of choice for FA patients that develop aplastic anemia, patients with a more stable bone marrow remain at a high risk of presenting MDS/AML and should be monitored for appearance of myeloid malignant clones. Markers for an as-early-as-possible identification of emerging myeloid malignant cells are needed for the monitoring of patients with FA, since quick medical action after detection of neoplastic transformation is needed.In this work we have leveraged publicly available single cell RNA seq (scRNAseq) datasets of patients with MDS and AML for training deep neural networks (DNN). We have generated two machine learning models aimed to identify myeloid malignant transcriptional profiles in scRNAseq datasets from the bone marrow of patients with FA, one for detection of MDS and a second one for AML. Both predictors displayed high sensitivity, specificity, and accuracy for detection of single cell resolution myeloid malignant transcriptional profiles.Multiple tools for analysis of single cell transcriptional data were implemented to characterize the predicted MDS and AML cells. Our analysis suggests that the predicted MDS and AML cells from FA patients are enriched in the lympho-myeloid-primed progenitor (LMPP) and the granulocyte-monocyte progenitor (GMP) populations. The predicted MDS and AML cells have gene expression and master transcriptional factor profiles that suggest malignant transformation and that differ from the rest of FA cells. Also cues of immune evasion were detected using single cell pathway analysis (SCPA) and cell-cell communication profiles. Next work will be aimed to find potential cell surface markers on the predicted MDS and AML cells as well as to assess our predictions in primary samples from FA patients.