The integration of computational models with experimental data is a cornerstone for gaining insight into biomedical applications. However, parameter fitting procedures often require a vast availability and frequency of data that are challenging to obtain from a single source. Here, we present a novel methodology called "CrossLabFit", which is designed to integrate data from multiple laboratories, overcoming the constraints of single-lab data collection. Our approach harmonizes disparate qualitative assessments, ranging from different experimental labs to categorical observations, into a unified framework for parameter estimation. By using machine learning clustering, these qualitative constraints are represented as dynamic "feasible windows" that capture significant trends to which models must adhere. For numerical implementation, we developed a GPU-accelerated version of differential evolution to navigate the cost function that integrated quantitative and qualitative information. We validate our approach across a series of case studies, demonstrating significant improvements in model accuracy and parameter identifiability. This work opens a new paradigm for collaborative science, enabling a methodological roadmap to combine and compare findings between studies to improve our understanding of biological systems and beyond.

In this paper we apply an inverse optimal controller (IOC) based on a control Lyapunov function (CLF) to schedule theoretical therapies for the novel coronavirus disease (COVID-19). This controller can represent the viral dynamics of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) in the host. The virus dynamics consider the antiviral effects and immune responses as control inputs. The proposed controller is based on a Recurrent High Order Neural Network (RHONN) used as an identifier trained with Extended Kalman Filter (EKF). Simulations show that applying treatment 2 days post symptoms would not significantly alter the viral load. The proposed controller to stimulate the immune response displays a better effectiveness compared to the effectiveness displayed by the antiviral effects.

MOTIVATION: Influenza viruses are a cause of large outbreaks and pandemics with high death tolls. A key obstacle is that flu vaccines have inconsistent performance, in the best cases up to 60% effectiveness, but it can be as low as 10%. Uncovering the hidden pathways of how antibodies (Abs) induced by one influenza strain are effective against another, cross-reaction, is a central vexation for the design of universal flu vaccines. RESULTS: We conceive a stochastic model that successfully represents the antibody cross-reactive data from mice infected with H3N2 influenza strains and further validation with cross-reaction data of H1N1 strains. Using a High-Performance Computing cluster, several aspects and parameters in the model were tested. Computational simulations highlight that changes in time of infection and the B-cells population are relevant, however, the affinity threshold of B-cells between consecutive infections is a necessary condition for the successful Abs cross-reaction. Our results suggest a 3-D reformulation of the current influenza antibody landscape for the representation and modeling of cross-reactive data. AVAILABILITY AND IMPLEMENTATION: The full code as a testing/simulation platform is freely available here: https://github.com/systemsmedicine/Antibody_cross-reaction_dynamics. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Partial differential equations (PDEs) is a well-established and powerful tool to simulate multi-cellular biological systems. However, available free tools for validation against data are on development. RESULTS: The PDEparams module provides a flexible interface and readily accommodates different parameter analysis tools in PDE models such as computation of likelihood profiles, and parametric bootstrapping, along with direct visualization of the results. To our knowledge, it is the first open, freely available tool for parameter fitting of PDE models. AVAILABILITY AND IMPLEMENTATION: PDEparams is distributed under the MIT license. The source code, usage instructions and examples are freely available on GitHub at github.com/systemsmedicine/PDE_params. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Infectious diseases are causes of morbidity and mortality worldwide. Mathematical models can serve as a central tool to predict the kinetic of different infections. However, the development of mechanistic models and their parameter estimation are difficult tasks. Using Recurrent High Order Neural Networks (RHONNs) trained with an algorithm based on the extended Kalman filter (EKF), we separately identified influenza A virus (IAV) and HIV dynamics. To this end, we considered within-host mathematical models of IAV and HIV as unknown signals to the RHONNs. Simulations results reported that for both infections, RHONNs are able to identify the within-host model dynamics. Results provide promising guidelines to tackle the problem of model identification of infectious diseases, serving for future model based control strategies of viral infections.

Parameter fitting consists on the estimation of model parameters using experimental data from the studied process, which can be considered as a nonlinear optimization problem. In this sense, evolutionary computation has shown its great capability to solve multimodal nonlinear optimization problems. This paper compares different variants of the Differential Evolution (DE) algorithm to minimize the residual sum of squares between the outcome of the mathematical model and experimental data. To compare the different variants of the DE algorithm, a biopolymer production model is considered. Simulations results suggest a trend for the best fit using the DE/best/ variants. However, the DE/rand/ variant provides more stable results respect to the average and standard deviation of different trials. Finally, the biopolymer production problem is discussed.

This paper presents a recurrent neural observer to estimate substrate and biomass concentrations in an activated sludge wastewater treatment. The observer is based on a discrete-time high order neural network (RHONN) trained on-line with an extended Kalman filter (EKF)-based algorithm. This observer is then associated with a hybrid intelligent system to control the substrate/biomass concentration ratio. The neural observer performance is illustrated via simulations.