Click here to read some interesting recently published papers from our community. If you have published an article in the field of in silico medicine, send it to us: we will include it in this section of the newsletter!
Journal of Royal Society Interface: Digital twinning of cardiac electrophysiology for congenital heart disease.
Matteo Salvador et al
Abstract
In recent years, blending mechanistic knowledge with machine learning has had a major impact in digital healthcare. In this work, we introduce a computational pipeline to build certified digital replicas of cardiac electrophysiology in paediatric patients with congenital heart disease. We construct the patient-specific geometry by means of semi-automatic segmentation and meshing tools. We generate a dataset of electrophysiology simulations covering cell-to-organ level model parameters and using rigorous mathematical models based on differential equations. We previously proposed Branched Latent Neural Maps (BLNMs) as an accurate and efficient means to recapitulate complex physical processes in a neural network. Here, we employ BLNMs to encode the parametrized temporal dynamics of in silico 12-lead electrocardiograms (ECGs). BLNMs act as a geometry-specific surrogate model of cardiac function for fast and robust parameter estimation to match clinical ECGs in paediatric patients. Identifiability and trustworthiness of calibrated model parameters are assessed by sensitivity analysis and uncertainty quantification.
---------------------------------------------------------
Frontiers in Digital Health: Toward mechanistic medical digital twins: some use cases in immunology.
Laubenbacher Reinhard et al
Abstract
The vision at the heart of personalized medicine is to design and implement customized approaches to keep individuals healthy and how to restore their health when it is compromised. This requires that we can quantify the differences between individuals that account for their health status in relation to their biological makeup, as well as the cumulative influences they are subjected to in their environment. Furthermore, we must also quantify their individual response to any therapeutic interventions over time. The only systematic way to accomplish this is to use computational models that can be personalized to an individual patient and can be dynamically recalibrated to reflect changes over time. Such models have become known as medical digital twins (MDTs). See Figure 1 for a workflow schematic of medical digital twin development. If we want to predict the individual response to treatment or develop novel drugs and other interventions, then these models need to be able to capture mechanisms and the effects of perturbing them. Given their use in patient treatment, requirements for model validation are much more stringent than for models used to discover new biology.
---------------------------------------------------------
Annals of Biomedical Engineering: Developing and Validating a Model of Humeral Stem Primary Stability, Intended for In Silico Clinical Trials.
Ghislain Maquer et al
Abstract
In silico clinical trials (ISCT) can contribute to demonstrating a device's performance via credible computational models applied on virtual cohorts. Our purpose was to establish the credibility of a model for assessing the risk of humeral stem loosening in total shoulder arthroplasty, based on a twofold validation scheme involving both benchtop and clinical validation activities, for ISCT applications. A finite element model computing bone-implant micromotion (benchtop model) was quantitatively compared to a bone foam micromotion test (benchtop comparator) to ensure that the physics of the system was captured correctly. The model was expanded to a population-based approach (clinical model) and qualitatively evaluated based on its ability to replicate findings from a published clinical study (clinical comparator), namely that grit-blasted stems are at a significantly higher risk of loosening than porous-coated stems, to ensure that clinical performance of the stem can be predicted appropriately. Model form sensitivities pertaining to surgical variation and implant design were evaluated. The model replicated benchtop micromotion measurements (52.1 ± 4.3 µm), without a significant impact of the press-fit ("Press-fit": 54.0 ± 8.5 µm, "No press-fit": 56.0 ± 12.0 µm). Applied to a virtual population, the grit-blasted stems (227 ± 78µm) experienced significantly larger micromotions than porous-coated stems (162 ± 69µm), in accordance with the findings of the clinical comparator. This work provides a concrete example for evaluating the credibility of an ISCT study. By validating the modeling approach against both benchtop and clinical data, model credibility is established for an ISCT application aiming to enrich clinical data in a regulatory submission.
---------------------------------------------------------
Toxicology and Applied Pharmacology: An in vitro-in silico workflow for predicting renal clearance of PFAS.
Hsing-Chieh Lin et al
Abstract
Per- and poly-fluoroalkyl substances (PFAS) have a wide range of elimination half-lives (days to years) in humans, thought to be in part due to variation in proximal tubule reabsorption. While human biomonitoring studies provide important data for some PFAS, renal clearance (CLrenal) predictions for hundreds of PFAS in commerce requires experimental studies with in vitro models and physiologically-based in vitro-to-in vivo extrapolation (IVIVE). Options for studying renal proximal tubule pharmacokinetics include cultures of renal proximal tubule epithelial cells (RPTECs) and/or microphysiological systems. This study aimed to compare CLrenal predictions for PFAS using in vitro models of varying complexity (96-well plates, static 24-well Transwells and a fluidic microphysiological model, all using human telomerase reverse transcriptase-immortalized and OAT1-overexpressing RPTECs combined with in silico physiologically-based IVIVE. Three PFAS were tested: one with a long half-life (PFOS) and two with shorter half-lives (PFHxA and PFBS). PFAS were added either individually (5 μM) or as a mixture (2 μM of each substance) for 48 h. Bayesian methods were used to fit concentrations measured in media and cells to a three-compartmental model to obtain the in vitro permeability rates, which were then used as inputs for a physiologically-based IVIVE model to estimate in vivo CLrenal. Our predictions for human CLrenal of PFAS were highly concordant with available values from in vivo human studies. The relative values of CLrenal between slow- and faster-clearance PFAS were most highly concordant between predictions from 2D culture and corresponding in vivo values. However, the predictions from the more complex model (with or without flow) exhibited greater concordance with absolute CLrenal. Overall, we conclude that a combined in vitro-in silico workflow can predict absolute CLrenal values, and effectively distinguish between PFAS with slow and faster clearance, thereby allowing prioritization of PFAS with a greater potential for bioaccumulation in humans.
---------------------------------------------------------
Clinical Pharmacology & Therapeutics: Simulating realistic patient profiles from pharmacokinetic models by a machine learning postprocessing correction of residual variability.
Christos Kaikousidis et al
Abstract
We address the problem of model misspecification in population pharmacokinetics (PopPK), by modeling residual unexplained variability (RUV) by machine learning (ML) methods in a postprocessing step after conventional model building. The practical purpose of the method is the generation of realistic virtual patient profiles and the quantification of the extent of model misspecification, by introducing an appropriate metric, to be used as an additional diagnostic of model quality. The proposed methodology consists of the following steps: After developing a PopPK model, the individual residual errors IRES = DV–IPRED, are computed, where DV are the observations and IPRED the individual predictions and are modeled by ML to obtain IRESML. Correction of the IPREDs can then be carried out as DVML= IPRED + IRESML. The methodology was tested in a PK study of ropinirole, for which a PopPK model was developed while a second deliberately misspecified model was also considered. Various supervised ML algorithms were tested, among which Random Forest gave the best results. The ML model was able to correct individual predictions as inspected in diagnostic plots and most importantly it simulated realistic profiles that resembled the real data and canceled out the artifacts introduced by the elevated RUV, even in the case of the heavily misspecified model. Furthermore, a metric to quantify the extent of model misspecification was introduced based on the R2 between IRES and IRESML, following the rationale that the greater the extent of variability explained by the ML model, the higher the degree of model misspecification present in the original model.
---------------------------------------------------------
