Generation of Virtual Patients for in Silico Cardiomyopathies Drug Development

Abstract

The revolution in modelling and simulation methodologies accompanied with the recent events in the highperformance computing (HPC) helped the development of in silico clinical platforms which integrate advanced and individualized simulation models to support drug development. These platforms incorporate patient specific models to create and generate virtual patients (VPs). A parametric methodology for resampling and generating VPs is the multivariate normal distribution which in the current work is optimized through an iterative pipeline by the Kolmogorov-Smirnov goodness-of-fit test. The proposed VP generator is integrated in the multirepository VP model of SILICOFCM which is a multi-modular, innovative in silico clinical trials solution for the design and functional optimization of the whole heart performance and monitoring effectiveness of pharmacological treatment for familial cardiomyopathies, with aim to reduce the animal studies and the human clinical trials.

Bayesian Inference-Based Gaussian Mixture Models With Optimal Components Estimation Towards Large-Scale Synthetic Data Generation for In Silico Clinical Trials

“…According to Supplementary Table I, the MVND and the log-MVND [4] , [5] , [36] are fast but they are based on critical assumptions (normality) and yield synthetic data with reduced quality. Although the BN offer explainable presentations of the conditional probabilities through the network, the different topologies are infinite, the quality of the virtual data is reduced, and the computational complexity is large [6] – [8] . The STE and UTE yield synthetic data with better quality, but they still have increased computational complexity for training/testing.…”

“…Four state-of-the art synthetic data generators [6] – [11] were used for comparison purposes, including the BN, the ANNs, the UTE, and the STE. Five quality indicators (Kullback-Leibler divergence, inter- and intra- correlation difference, goodness of fit - GOF, coefficient of variation - cV) [29] – [31] were used to measure the similarity, dispersity, and divergence between the synthetic and the real data (Supplementary Material, Section F).…”

“…The MVND was applied in [4] , [5] to generate virtual data based on the mean and the covariance of the real data. In addition, the BN were used in [6] – [8] for the generation of synthetic distributions based on the modeling of conditional probabilities across diverse network topologies. The BN and the MVND, however, suffer from mathematical assumptions; the MVND algorithm assumes that the real data are normally distributed whereas in the BNs the conditional probabilities are modeled using assumptions on the prior distribution of the features.…”

Variational Gaussian Mixture Models with robust Dirichlet concentration priors for virtual population generation in hypertrophic cardiomyopathy: a comparison study

“…Probabilistic approaches are the most common methods for virtual population generation [5,6], where the synthetic samples are randomly drawn from the real distributions. A widely used probabilistic method is the multivariate normal distribution (MVND) which has been widely adopted for the generation of virtual patients in in-silico clinical trials, such as, in hypertrophic cardiomyopathy (HCM) [6].…”

Generation of virtual patient data for in-silico cardiomyopathies drug development using tree ensembles: a comparative study

“…In this work, we deploy three computational methods to generate virtual patient data from real clinical data. We extend a previous study [6], where the MVND method was used to generate 300 virtual patients by comparing the multivariate log-normal distribution with the tree ensembles to generate 1000 virtual patients for in-silico clinical trials targeting the drug development for familiar cardiomyopathies (FCM). These methods have been integrated into the in-silico clinical trial SILICOFCM cloud-based platform [7].…”