Predicting in vitro human mesenchymal stromal cell expansion based on individual donor characteristics using machine learning.

BACKGROUND: Human mesenchymal stromal cells (hMSCs) have become attractive candidates for advanced medical cell-based therapies. An in vitro expansion step is routinely used to reach the required clinical quantities. However, this is influenced by many variables including donor characteristics, such as age and gender, and culture conditions, such as cell seeding density and available culture surface area. Computational modeling in general and machine learning in particular could play a significant role in deciphering the relationship between the individual donor characteristics and their growth dynamics. METHODS: In this study, hMSCs obtained from 174 male and female donors, between 3 and 64 years of age with passage numbers ranging from 2 to 27, were studied. We applied a Random Forests (RF) technique to model the cell expansion procedure by predicting the population doubling time (PDT) for each passage, taking into account individual donor-related characteristics. RESULTS: Using the RF model, the mean absolute error between model predictions and experimental results for the PDT in passage 1 to 4 is significantly lower compared with the errors obtained with theoretical estimates or historical data. Moreover, statistical analysis indicate that the PD and PDT in different age categories are significantly different, especially in the youngest group (younger than 10 years of age) compared with the other age groups. DISCUSSION: In summary, we introduce a predictive computational model describing in vitro cell expansion dynamics based on individual donor characteristics, an approach that could greatly assist toward automation of a cell expansion culture process.

Maximizing neotissue growth kinetics in a perfusion bioreactor: An in silico strategy using model reduction and Bayesian optimization.

In regenerative medicine, computer models describing bioreactor processes can assist in designing optimal process conditions leading to robust and economically viable products. In this study, we started from a (3D) mechanistic model describing the growth of neotissue, comprised of cells, and extracellular matrix, in a perfusion bioreactor set-up influenced by the scaffold geometry, flow-induced shear stress, and a number of metabolic factors. Subsequently, we applied model reduction by reformulating the problem from a set of partial differential equations into a set of ordinary differential equations. Comparing the reduced model results to the mechanistic model results and to dedicated experimental results assesses the reduction step quality. The obtained homogenized model is 105 fold faster than the 3D version, allowing the application of rigorous optimization techniques. Bayesian optimization was applied to find the medium refreshment regime in terms of frequency and percentage of medium replaced that would maximize neotissue growth kinetics during 21 days of culture. The simulation results indicated that maximum neotissue growth will occur for a high frequency and medium replacement percentage, a finding that is corroborated by reports in the literature. This study demonstrates an in silico strategy for bioprocess optimization paying particular attention to the reduction of the associated computational cost.

Optimizing neotissue growth inside perfusion bioreactors with respect to culture and labor cost: a multi-objective optimization study using evolutionary algorithms.

Tissue engineering is a fast progressing domain where solutions are provided for organ failure or tissue damage. In this domain, computer models can facilitate the design of optimal production process conditions leading to robust and economically viable products. In this study, we use a previously published computationally efficient model, describing the neotissue growth (cells + their extracellular matrix) inside 3D scaffolds in a perfusion bioreactor. In order to find the most cost-effective medium refreshment strategy for the bioreactor culture, a multi-objective optimization strategy was developed aimed at maximizing the neotissue growth while minimizing the total cost of the experiment. Four evolutionary optimization algorithms (NSGAII, MOPSO, MOEA/D and GDEIII) were applied to the problem and the Pareto frontier was computed in all methods. All algorithms led to a similar outcome, albeit with different convergence speeds. The simulation results indicated that, given the actual cost of the labor compared to the medium cost, the most cost-efficient way of refreshing the medium was obtained by minimizing the refreshment frequency and maximizing the refreshment amount.

Computational Modeling of Human Mesenchymal Stromal Cell Proliferation and Extra-Cellular Matrix Production in 3D Porous Scaffolds in a Perfusion Bioreactor: The Effect of Growth Factors.

Stem cell expansion on 3D porous scaffolds cultured in bioreactor systems has been shown to be beneficial for maintenance of the original cell functionality in tissue engineering strategies (TE). However, the production of extracellular matrix (ECM) makes harvesting the progenitor cell population from 3D scaffolds a challenge. Medium composition plays a role in stimulating cell proliferation over extracellular matrix (ECM) production. In this regard, a computational model describing tissue growth inside 3D scaffolds can be a great tool in designing optimal experimental conditions. In this study, a computational model describing cell and ECM growth in a perfusion bioreactor is developed, including a description of the effect of a (generic) growth factor on the biological processes taking place inside the 3D scaffold. In the model, the speed of cell and ECM growth depends on the flow-induced shear stress, curvature and the concentrations of oxygen, glucose, lactate, and growth factor. The effect of the simulated growth factor is to differentially enhance cell proliferation over ECM production. After model calibration with historic in-house data, a multi-objective optimization procedure is executed aiming to minimize the total experimental cost whilst maximizing cell growth during culture. The obtained results indicate there are multiple optimum points for the medium refreshment regime and the initial growth factor concentration where a trade-off is made between the final amount of cells and the culture cost. Finally, the model is applied to experiments reported in the literature studying the effects of perfusion-based cell culture and/or growth factor supplementation on cell expansion. The qualitative similarities between the simulation and experimental results, even in the absence of proper model calibration, reinforces the generic character of the proposed modeling framework. The model proposed in this study can contribute to the cost efficient production of cell-based TE products, ultimately contributing to their affordability and accessibility.

Bayesian Multiobjective Optimisation With Mixed Analytical and Black-Box Functions: Application to Tissue Engineering.

Tissue engineering and regenerative medicine looks at improving or restoring biological tissue function in humans and animals. We consider optimising neotissue growth in a three-dimensional scaffold during dynamic perfusion bioreactor culture, in the context of bone tissue engineering. The goal is to choose design variables that optimise two conflicting objectives, first, maximising neotissue growth and, second, minimising operating cost. We make novel extensions to Bayesian multiobjective optimisation in the case of one analytical objective function and one black-box, i.e. simulation based and objective function. The analytical objective represents operating cost while the black-box neotissue growth objective comes from simulating a system of partial differential equations. The resulting multiobjective optimisation method determines the tradeoff between neotissue growth and operating cost. Our method exhibits better data efficiency than genetic algorithms, i.e. the most common approach in the literature, on both the tissue engineering example and standard test functions. The multiobjective optimisation method applies to real-world problems combining black-box models with easy-to-quantify objectives such as cost.

Modeling of tumor growth in dendritic cell-based immunotherapy using artificial neural networks.

Exposure-response modeling and simulation is especially useful in oncology as it permits to predict and design un-experimented clinical trials as well as dose selection. Dendritic cells (DC) are the most effective immune cells in the regulation of immune system. To activate immune system, DCs may be matured by many factors like bacterial CpG-DNA, Lipopolysaccharaide (LPS) and other microbial products. In this paper, a model based on artificial neural network (ANN) is presented for analyzing the dynamics of antitumor vaccines using empirical data obtained from the experimentations of different groups of mice treated with DCs matured by bacterial CpG-DNA, LPS and whole lysate of a Gram-positive bacteria Listeria monocytogenes. Also, tumor lysate was added to DCs followed by addition of maturation factors. Simulations show that the proposed model can interpret the important features of empirical data. Owing to the nonlinearity properties, the proposed ANN model has been able not only to describe the contradictory empirical results, but also to predict new vaccination patterns for controlling the tumor growth. For example, the proposed model predicts an exponentially increasing pattern of CpG-matured DC to be effective in suppressing the tumor growth.