Accounting for uncertainty in training data to improve machine learning performance in predicting new disease activity in early multiple sclerosis.

Introduction: Machine learning (ML) has great potential for using health data to predict clinical outcomes in individual patients. Missing data are a common challenge in training ML algorithms, such as when subjects withdraw from a clinical study, leaving some samples with missing outcome labels. In this study, we have compared three ML models to determine whether accounting for label uncertainty can improve a model's predictions. Methods: We used a dataset from a completed phase-III clinical trial that evaluated the efficacy of minocycline for delaying the conversion from clinically isolated syndrome to multiple sclerosis (MS), using the McDonald 2005 diagnostic criteria. There were a total of 142 participants, and at the 2-year follow-up 81 had converted to MS, 29 remained stable, and 32 had uncertain outcomes. In a stratified 7-fold cross-validation, we trained three random forest (RF) ML models using MRI volumetric features and clinical variables to predict the conversion outcome, which represented new disease activity within 2 years of a first clinical demyelinating event. One RF was trained using subjects with the uncertain labels excluded (RFexclude), another RF was trained using the entire dataset but with assumed labels for the uncertain group (RFnaive), and a third, a probabilistic RF (PRF, a type of RF that can model label uncertainty) was trained on the entire dataset, with probabilistic labels assigned to the uncertain group. Results: Probabilistic random forest outperformed both the RF models with the highest AUC (0.76, compared to 0.69 for RFexclude and 0.71 for RFnaive) and F1-score (86.6% compared to 82.6% for RFexclude and 76.8% for RFnaive). Conclusion: Machine learning algorithms capable of modeling label uncertainty can improve predictive performance in datasets in which a substantial number of subjects have unknown outcomes.

A distinct microglial subset at the tumor-stroma interface of glioma.

The characterization of the tumor microenvironment (TME) in high grade gliomas (HGG) has generated significant interest in an effort to understand how neoplastic lesions in the central nervous system (CNS) are supported and to devise novel therapeutic targets. The TME of the CNS contains unique and specialized cells, including the resident myeloid cells, microglia. Myeloid involvement in HGG, such as glioblastoma, is associated with poor outcomes. Glioma-associated microglia and infiltrating monocytes/macrophages (GAM) accumulate within the neoplastic lesion where they facilitate tumor growth and drive immunosuppression. However, it has been difficult to differentiate whether microglia and macrophages have similar or distinct roles in pathology, and if the spatial organization of these cells informs outcomes. Here, we characterize the tumor-stroma border and identify peritumoral GAM (PGAM) as a unique subpopulation of GAM. Using data mining and analyses of samples derived from both murine and human sources we show that PGAM exhibit a pro-inflammatory and chemotactic phenotype that is associated with peripheral monocyte recruitment, and decreased overall survival. PGAM act as a unique subset of GAM at the tumor-stroma interface. We define a novel gene signature to identify these cells and suggest that PGAM constitute a cellular target of the TME.

A Rigorous Quantitative Approach to Analyzing Phagocytosis Assays.

Studying monocytic cells in isolated systems in vitro contributes significantly to the understanding of innate immune physiology. Functional assays produce read outs which can be used to measure responses to selected stimuli, such as pathogen exposure, antigen loading, and cytokine stimulation. Integration of these results with high quality in vivo models allows for the development of therapeutics which target these cell populations. Current methodologies to quantify phagocytic function of monocytic cells in vitro either measure phagocytic activity of individual cells (average number of beads or particles/cell), or a population outcome (% cells that contain phagocytosed material). Here we address technical challenges and shortcomings of these methods and present a protocol for collecting and analyzing data derived from a functional assay which measures phagocytic activity of macrophage and macrophage-like cells. We apply this method to two different experimental conditions, and compare to existing work flows. We also provide an online tool for users to upload and analyze data using this method.