Single-cell transcriptomics has revolutionized our understanding of cellular heterogeneity and drug perturbation effects. However, its high cost and the vast chemical space of potential drugs present barriers to experimentally characterizing the effect of chemical perturbations in all the myriad cell types of the human body. To overcome these limitations, several groups have proposed using machine learning methods to directly predict the effect of chemical perturbations either across cell contexts or chemical space. However, advances in this field have been hindered by a lack of well-designed evaluation datasets and benchmarks. To drive innovation in perturbation modeling, the Open Problems Perturbation Prediction (OP3) benchmark introduces a framework for predicting the effects of small molecule perturbations on cell type-specific gene expression. OP3 leverages the Open Problems in Single-cell Analysis benchmarking infrastructure and is enabled by a new single-cell perturbation dataset, encompassing 146 compounds tested on human blood cells. The benchmark includes diverse data representations, evaluation metrics, and winning methods from our "Single-cell perturbation prediction: generalizing experimental interventions to unseen contexts" competition at NeurIPS 2023. We envision that the OP3 benchmark and competition will drive innovation in single-cell perturbation prediction by improving the accessibility, visibility, and feasibility of this challenge, thereby promoting the impact of machine learning in drug discovery.

Magnesium (Mg2+) ions are frequently reported to regulate osteogenic activities of mesenchymal stem cells (MSCs). In this study, we propose a numerical model to study the regulatory importance of Mg2+ ions on MSCs osteoblastic differentiation in the presence of an inflammatory response. A fuzzy logic controller was formulated to receive the concentrations of Mg2+ ions and the inflammatory cytokines of TNF-α, IL-10, IL-1β, and IL-8 as cellular inputs and predict the cells' early and late differentiation rates. Five sets of empirical data obtained from published cell culture experiments were used to calibrate the model. The model successfully reproduced the empirical data regarding the concentration- and phase-dependent effect of Mg2+ ions on the differentiation process. In agreement with the experiments, the model showed the stimulatory role of Mg2+ ions on the early differentiation phase, once administered at low concentration, and their inhibitory role on the late differentiation phase. The numerical approach used in this study suggested 6-8 mM as the most effective concentration of Mg2+ ions in promoting the early differentiation process. Also, the proposed model sheds light on the fundamental differences in the behavioral properties of cells cultured in different experiments, e.g. differentiation rate and the sensitivity of the cultured cells to stimulatory signals such as Mg2+ ions. Thus, it can be used to interpret and compare different empirical findings. Moreover, the model successfully reproduced the nonlinearities in the concentration-dependent role of the inflammatory cytokines in early and late differentiation rates. Overall, the proposed model can be employed in studying the osteogenic properties of Mg-based implants in the presence of an inflammatory response.

Abstract Agent‐based modeling (ABM) has been extensively used to study the collective behavior of systems emerging from the interaction of numerous independent individuals called agents. Python and C++ are commonly used for ABM thanks to their unique features; the latter offers superior performance while the former provides ease‐of‐use and rich libraries in data science, visualization, and machine learning. We present the framework CppyABM that unifies these features by providing identical ABM semantic and development styles in both C++ and Python as well as the essential binding tools to expose a certain functionality from C++ to Python. The binding feature allows users to tailor and further extend a type or function within Python while it is originally defined in C++. Using CppyABM, users can choose either C++ or Python depending on their expertise and the specialty of the model or combine them to benefit from the advantages of both languages simultaneously. We provide showcases of CppyABM capabilities using several examples in computational biology, ecology, and virology. These examples are implemented in different formats using either C++ or Python or a combination of both to provide a comparison between the performance of implementation scenarios. The results of the example show a clear performance advantage of the models entirely or partly implemented in C++ compared to purely Python‐based implementations.