A call for virtual experiments: accelerating the scientific process

Engineer design process assisted by explainable deep learning network

Engineering simulation accelerates the development of reliable and repeatable design processes in various domains. However, the computing resource consumption is dramatically raised in the whole development processes. Making the most of these simulation data becomes more and more important in modern industrial product design. In the present study, we proposed a workflow comprised of a series of machine learning algorithms (mainly deep neuron networks) to be an alternative to the numerical simulation. We have applied the workflow to the field of dental implant design process. The process is based on a complex, time-dependent, multi-physical biomechanical theory, known as mechano-regulatory method. It has been used to evaluate the performance of dental implants and to assess the tissue recovery after the oral surgery procedures. We provided a deep learning network (DLN) with calibrated simulation data that came from different simulation conditions with experimental verification. The DLN achieves nearly exact result of simulated bone healing history around implants. The correlation of the predicted essential physical properties of surrounding bones (e.g. strain and fluid velocity) and performance indexes of implants (e.g. bone area and bone-implant contact) were greater than 0.980 and 0.947, respectively. The testing AUC values for the classification of each tissue phenotype were ranging from 0.90 to 0.99. The DLN reduced hours of simulation time to seconds. Moreover, our DLN is explainable via Deep Taylor decomposition, suggesting that the transverse fluid velocity, upper and lower parts of dental implants are the keys that influence bone healing and the distribution of tissue phenotypes the most. Many examples of commercial dental implants with designs which follow these design strategies can be found. This work demonstrates that DLN with proper network design is capable to replace complex, time-dependent, multi-physical models/theories, as well as to reveal the underlying features without prior professional knowledge.

Towards standardization guidelines for in silico approaches in personalized medicine

Despite the ever-progressing technological advances in producing data in health and clinical research, the generation of new knowledge for medical benefits through advanced analytics still lags behind its full potential. Reasons for this obstacle are the inherent heterogeneity of data sources and the lack of broadly accepted standards. Further hurdles are associated with legal and ethical issues surrounding the use of personal/patient data across disciplines and borders. Consequently, there is a need for broadly applicable standards compliant with legal and ethical regulations that allow interpretation of heterogeneous health data through in silico methodologies to advance personalized medicine. To tackle these standardization challenges, the Horizon2020 Coordinating and Support Action EU-STANDS4PM initiated an EU-wide mapping process to evaluate strategies for data integration and data-driven in silico modelling approaches to develop standards, recommendations and guidelines for personalized medicine. A first step towards this goal is a broad stakeholder consultation process initiated by an EU-STANDS4PM workshop at the annual COMBINE meeting (COMBINE 2019 workshop report in same issue). This forum analysed the status quo of data and model standards and reflected on possibilities as well as challenges for cross-domain data integration to facilitate in silico modelling approaches for personalized medicine.

Calibration of ionic and cellular cardiac electrophysiology models

Cardiac electrophysiology models are among the most mature and well-studied mathematical models of biological systems. This maturity is bringing new challenges as models are being used increasingly to make quantitative rather than qualitative predictions. As such, calibrating the parameters within ion current and action potential (AP) models to experimental data sets is a crucial step in constructing a predictive model. This review highlights some of the fundamental concepts in cardiac model calibration and is intended to be readily understood by computational and mathematical modelers working in other fields of biology. We discuss the classic and latest approaches to calibration in the electrophysiology field, at both the ion channel and cellular AP scales. We end with a discussion of the many challenges that work to date has raised and the need for reproducible descriptions of the calibration process to enable models to be recalibrated to new data sets and built upon for new studies. This article is categorized under: Analytical and Computational Methods > Computational Methods Physiology > Mammalian Physiology in Health and Disease Models of Systems Properties and Processes > Cellular Models.

Towards reproducible computational drug discovery

The reproducibility of experiments has been a long standing impediment for further scientific progress. Computational methods have been instrumental in drug discovery efforts owing to its multifaceted utilization for data collection, pre-processing, analysis and inference. This article provides an in-depth coverage on the reproducibility of computational drug discovery. This review explores the following topics: (1) the current state-of-the-art on reproducible research, (2) research documentation (e.g. electronic laboratory notebook, Jupyter notebook, etc.), (3) science of reproducible research (i.e. comparison and contrast with related concepts as replicability, reusability and reliability), (4) model development in computational drug discovery, (5) computational issues on model development and deployment, (6) use case scenarios for streamlining the computational drug discovery protocol. In computational disciplines, it has become common practice to share data and programming codes used for numerical calculations as to not only facilitate reproducibility, but also to foster collaborations (i.e. to drive the project further by introducing new ideas, growing the data, augmenting the code, etc.). It is therefore inevitable that the field of computational drug design would adopt an open approach towards the collection, curation and sharing of data/code.

Gsmodutils: a python based framework for test-driven genome scale metabolic model development

MOTIVATION

Genome scale metabolic models (GSMMs) are increasingly important for systems biology and metabolic engineering research as they are capable of simulating complex steady-state behaviour. Constraints based models of this form can include thousands of reactions and metabolites, with many crucial pathways that only become activated in specific simulation settings. However, despite their widespread use, power and the availability of tools to aid with the construction and analysis of large scale models, little methodology is suggested for their continued management. For example, when genome annotations are updated or new understanding regarding behaviour is discovered, models often need to be altered to reflect this. This is quickly becoming an issue for industrial systems and synthetic biotechnology applications, which require good quality reusable models integral to the design, build, test and learn cycle.

RESULTS

As part of an ongoing effort to improve genome scale metabolic analysis, we have developed a test-driven development methodology for the continuous integration of validation data from different sources. Contributing to the open source technology based around COBRApy, we have developed the gsmodutils modelling framework placing an emphasis on test-driven design of models through defined test cases. Crucially, different conditions are configurable allowing users to examine how different designs or curation impact a wide range of system behaviours, minimizing error between model versions.

AVAILABILITY AND IMPLEMENTATION

The software framework described within this paper is open source and freely available from http://github.com/SBRCNottingham/gsmodutils.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Harmonizing semantic annotations for computational models in biology

Life science researchers use computational models to articulate and test hypotheses about the behavior of biological systems. Semantic annotation is a critical component for enhancing the interoperability and reusability of such models as well as for the integration of the data needed for model parameterization and validation. Encoded as machine-readable links to knowledge resource terms, semantic annotations describe the computational or biological meaning of what models and data represent. These annotations help researchers find and repurpose models, accelerate model composition and enable knowledge integration across model repositories and experimental data stores. However, realizing the potential benefits of semantic annotation requires the development of model annotation standards that adhere to a community-based annotation protocol. Without such standards, tool developers must account for a variety of annotation formats and approaches, a situation that can become prohibitively cumbersome and which can defeat the purpose of linking model elements to controlled knowledge resource terms. Currently, no consensus protocol for semantic annotation exists among the larger biological modeling community. Here, we report on the landscape of current annotation practices among the COmputational Modeling in BIology NEtwork community and provide a set of recommendations for building a consensus approach to semantic annotation.

Data Management in Computational Systems Biology: Exploring Standards, Tools, Databases, and Packaging Best Practices

Computational systems biology involves integrating heterogeneous datasets in order to generate models. These models can assist with understanding and prediction of biological phenomena. Generating datasets and integrating them into models involves a wide range of scientific expertise. As a result these datasets are often collected by one set of researchers, and exchanged with others researchers for constructing the models. For this process to run smoothly the data and models must be FAIR-findable, accessible, interoperable, and reusable. In order for data and models to be FAIR they must be structured in consistent and predictable ways, and described sufficiently for other researchers to understand them. Furthermore, these data and models must be shared with other researchers, with appropriately controlled sharing permissions, before and after publication. In this chapter we explore the different data and model standards that assist with structuring, describing, and sharing. We also highlight the popular standards and sharing databases within computational systems biology.

Reproducible model development in the cardiac electrophysiology Web Lab

The modelling of the electrophysiology of cardiac cells is one of the most mature areas of systems biology. This extended concentration of research effort brings with it new challenges, foremost among which is that of choosing which of these models is most suitable for addressing a particular scientific question. In a previous paper, we presented our initial work in developing an online resource for the characterisation and comparison of electrophysiological cell models in a wide range of experimental scenarios. In that work, we described how we had developed a novel protocol language that allowed us to separate the details of the mathematical model (the majority of cardiac cell models take the form of ordinary differential equations) from the experimental protocol being simulated. We developed a fully-open online repository (which we termed the Cardiac Electrophysiology Web Lab) which allows users to store and compare the results of applying the same experimental protocol to competing models. In the current paper we describe the most recent and planned extensions of this work, focused on supporting the process of model building from experimental data. We outline the necessary work to develop a machine-readable language to describe the process of inferring parameters from wet lab datasets, and illustrate our approach through a detailed example of fitting a model of the hERG channel using experimental data. We conclude by discussing the future challenges in making further progress in this domain towards our goal of facilitating a fully reproducible approach to the development of cardiac cell models.

The Cardiac Electrophysiology Web Lab

Computational modeling of cardiac cellular electrophysiology has a long history, and many models are now available for different species, cell types, and experimental preparations. This success brings with it a challenge: how do we assess and compare the underlying hypotheses and emergent behaviors so that we can choose a model as a suitable basis for a new study or to characterize how a particular model behaves in different scenarios? We have created an online resource for the characterization and comparison of electrophysiological cell models in a wide range of experimental scenarios. The details of the mathematical model (quantitative assumptions and hypotheses formulated as ordinary differential equations) are separated from the experimental protocol being simulated. Each model and protocol is then encoded in computer-readable formats. A simulation tool runs virtual experiments on models encoded in CellML, and a website (https://chaste.cs.ox.ac.uk/WebLab) provides a friendly interface, allowing users to store and compare results. The system currently contains a sample of 36 models and 23 protocols, including current-voltage curve generation, action potential properties under steady pacing at different rates, restitution properties, block of particular channels, and hypo-/hyperkalemia. This resource is publicly available, open source, and free, and we invite the community to use it and become involved in future developments. Investigators interested in comparing competing hypotheses using models can make a more informed decision, and those developing new models can upload them for easy evaluation under the existing protocols, and even add their own protocols.

Reproducibility in Computational Neuroscience Models and Simulations

OBJECTIVE

Like all scientific research, computational neuroscience research must be reproducible. Big data science, including simulation research, cannot depend exclusively on journal articles as the method to provide the sharing and transparency required for reproducibility.

METHODS

Ensuring model reproducibility requires the use of multiple standard software practices and tools, including version control, strong commenting and documentation, and code modularity.

RESULTS

Building on these standard practices, model-sharing sites and tools have been developed that fit into several categories: 1) standardized neural simulators; 2) shared computational resources; 3) declarative model descriptors, ontologies, and standardized annotations; and 4) model-sharing repositories and sharing standards.

CONCLUSION

A number of complementary innovations have been proposed to enhance sharing, transparency, and reproducibility. The individual user can be encouraged to make use of version control, commenting, documentation, and modularity in development of models. The community can help by requiring model sharing as a condition of publication and funding.

SIGNIFICANCE

Model management will become increasingly important as multiscale models become larger, more detailed, and correspondingly more difficult to manage by any single investigator or single laboratory. Additional big data management complexity will come as the models become more useful in interpreting experiments, thus increasing the need to ensure clear alignment between modeling data, both parameters and results, and experiment.

Where next for the reproducibility agenda in computational biology?

Background

The concept of reproducibility is a foundation of the scientific method. With the arrival of fast and powerful computers over the last few decades, there has been an explosion of results based on complex computational analyses and simulations. The reproducibility of these results has been addressed mainly in terms of exact replicability or numerical equivalence, ignoring the wider issue of the reproducibility of conclusions through equivalent, extended or alternative methods.

Results

We use case studies from our own research experience to illustrate how concepts of reproducibility might be applied in computational biology. Several fields have developed ‘minimum information’ checklists to support the full reporting of computational simulations, analyses and results, and standardised data formats and model description languages can facilitate the use of multiple systems to address the same research question. We note the importance of defining the key features of a result to be reproduced, and the expected agreement between original and subsequent results. Dynamic, updatable tools for publishing methods and results are becoming increasingly common, but sometimes come at the cost of clear communication. In general, the reproducibility of computational research is improving but would benefit from additional resources and incentives.

Conclusions

We conclude with a series of linked recommendations for improving reproducibility in computational biology through communication, policy, education and research practice. More reproducible research will lead to higher quality conclusions, deeper understanding and more valuable knowledge.

The Human Physiome: how standards, software and innovative service infrastructures are providing the building blocks to make it achievable

Reconstructing and understanding the Human Physiome virtually is a complex mathematical problem, and a highly demanding computational challenge. Mathematical models spanning from the molecular level through to whole populations of individuals must be integrated, then personalized. This requires interoperability with multiple disparate and geographically separated data sources, and myriad computational software tools. Extracting and producing knowledge from such sources, even when the databases and software are readily available, is a challenging task. Despite the difficulties, researchers must frequently perform these tasks so that available knowledge can be continually integrated into the common framework required to realize the Human Physiome. Software and infrastructures that support the communities that generate these, together with their underlying standards to format, describe and interlink the corresponding data and computer models, are pivotal to the Human Physiome being realized. They provide the foundations for integrating, exchanging and re-using data and models efficiently, and correctly, while also supporting the dissemination of growing knowledge in these forms. In this paper, we explore the standards, software tooling, repositories and infrastructures that support this work, and detail what makes them vital to realizing the Human Physiome.

Myokit: A simple interface to cardiac cellular electrophysiology

Myokit is a new powerful and versatile software tool for modeling and simulation of cardiac cellular electrophysiology. Myokit consists of an easy-to-read modeling language, a graphical user interface, single and multi-cell simulation engines and a library of advanced analysis tools accessible through a Python interface. Models can be loaded from Myokit's native file format or imported from CellML. Model export is provided to C, MATLAB, CellML, CUDA and OpenCL. Patch-clamp data can be imported and used to estimate model parameters. In this paper, we review existing tools to simulate the cardiac cellular action potential to find that current tools do not cater specifically to model development and that there is a gap between easy-to-use but limited software and powerful tools that require strong programming skills from their users. We then describe Myokit's capabilities, focusing on its model description language, simulation engines and import/export facilities in detail. Using three examples, we show how Myokit can be used for clinically relevant investigations, multi-model testing and parameter estimation in Markov models, all with minimal programming effort from the user. This way, Myokit bridges a gap between performance, versatility and user-friendliness.

MioLab, a rat cardiac contractile force simulator: Applications to teaching cardiac cell physiology and biophysics

INTRODUCTION

Understanding the basic concepts of physiology and biophysics of cardiac cells can be improved by virtual experiments that illustrate the complex excitation-contraction coupling process in cardiac cells. The aim of this study is to propose a rat cardiac myocyte simulator, with which calcium dynamics in excitation-contraction coupling of an isolated cell can be observed. This model has been used in the course "Mathematical Modeling and Simulation of Biological Systems". In this paper we present the didactic utility of the simulator MioLab(®).

METHODS

The simulator enables virtual experiments that can help studying inhibitors and activators in the sarcoplasmic reticulum sodium-calcium exchanger, thus corroborating a better understanding of the effects of medications, which are used to treat arrhythmias, on these compartments. The graphical interfaces were developed not only to facilitate the use of the simulator, but also to promote a constructive learning on the subject, since there are animations and videos for each stage of the simulation. The effectiveness of the simulator was tested by a group of graduate students.

RESULTS

Some examples of simulations were presented in order to describe the overall structure of the simulator. Part of these virtual experiments became an activity for Biomedical Engineering graduate students, who evaluated the simulator based on its didactic quality. As a result, students answered a questionnaire on the usability and functionality of the simulator as a teaching tool. All students performed the proposed activities and classified the simulator as an optimal or good learning tool. In their written questions, students indicated as negative characteristics some problems with visualizing graphs; as positive characteristics, they indicated the simulator's didactic function, especially tutorials and videos on the topic of this study.

CONCLUSIONS

The results show that the simulator complements the study of the physiology and biophysics of the cardiac cell.

Hodgkin-Huxley revisited: reparametrization and identifiability analysis of the classic action potential model with approximate Bayesian methods

As cardiac cell models become increasingly complex, a correspondingly complex 'genealogy' of inherited parameter values has also emerged. The result has been the loss of a direct link between model parameters and experimental data, limiting both reproducibility and the ability to re-fit to new data. We examine the ability of approximate Bayesian computation (ABC) to infer parameter distributions in the seminal action potential model of Hodgkin and Huxley, for which an immediate and documented connection to experimental results exists. The ability of ABC to produce tight posteriors around the reported values for the gating rates of sodium and potassium ion channels validates the precision of this early work, while the highly variable posteriors around certain voltage dependency parameters suggests that voltage clamp experiments alone are insufficient to constrain the full model. Despite this, Hodgkin and Huxley's estimates are shown to be competitive with those produced by ABC, and the variable behaviour of posterior parametrized models under complex voltage protocols suggests that with additional data the model could be fully constrained. This work will provide the starting point for a full identifiability analysis of commonly used cardiac models, as well as a template for informative, data-driven parametrization of newly proposed models.

Combining computational models, semantic annotations and simulation experiments in a graph database

Model repositories such as the BioModels Database, the CellML Model Repository or JWS Online are frequently accessed to retrieve computational models of biological systems. However, their storage concepts support only restricted types of queries and not all data inside the repositories can be retrieved. In this article we present a storage concept that meets this challenge. It grounds on a graph database, reflects the models' structure, incorporates semantic annotations and simulation descriptions and ultimately connects different types of model-related data. The connections between heterogeneous model-related data and bio-ontologies enable efficient search via biological facts and grant access to new model features. The introduced concept notably improves the access of computational models and associated simulations in a model repository. This has positive effects on tasks such as model search, retrieval, ranking, matching and filtering. Furthermore, our work for the first time enables CellML- and Systems Biology Markup Language-encoded models to be effectively maintained in one database. We show how these models can be linked via annotations and queried. Database URL: https://sems.uni-rostock.de/projects/masymos/

OpenCOR: a modular and interoperable approach to computational biology

Computational biologists have been developing standards and formats for nearly two decades, with the aim of easing the description and exchange of experimental data, mathematical models, simulation experiments, etc. One of those efforts is CellML (cellml.org), an XML-based markup language for the encoding of mathematical models. Early CellML-based environments include COR and OpenCell. However, both of those tools have limitations and were eventually replaced with OpenCOR (opencor.ws). OpenCOR is an open source modeling environment that is supported on Windows, Linux and OS X. It relies on a modular approach, which means that all of its features come in the form of plugins. Those plugins can be used to organize, edit, simulate and analyze models encoded in the CellML format. We start with an introduction to CellML and two of its early adopters, which limitations eventually led to the development of OpenCOR. We then go onto describing the general philosophy behind OpenCOR, as well as describing its openness and its development process. Next, we illustrate various aspects of OpenCOR, such as its user interface and some of the plugins that come bundled with it (e.g., its editing and simulation plugins). Finally, we discuss some of the advantages and limitations of OpenCOR before drawing some concluding remarks.

Cellular cardiac electrophysiology modeling with Chaste and CellML

Chaste is an open-source C++ library for computational biology that has well-developed cardiac electrophysiology tissue simulation support. In this paper, we introduce the features available for performing cardiac electrophysiology action potential simulations using a wide range of models from the Physiome repository. The mathematics of the models are described in CellML, with units for all quantities. The primary idea is that the model is defined in one place (the CellML file), and all model code is auto-generated at compile or run time; it never has to be manually edited. We use ontological annotation to identify model variables describing certain biological quantities (membrane voltage, capacitance, etc.) to allow us to import any relevant CellML models into the Chaste framework in consistent units and to interact with them via consistent interfaces. This approach provides a great deal of flexibility for analysing different models of the same system. Chaste provides a wide choice of numerical methods for solving the ordinary differential equations that describe the models. Fixed-timestep explicit and implicit solvers are provided, as discussed in previous work. Here we introduce the Rush–Larsen and Generalized Rush–Larsen integration techniques, made available via symbolic manipulation of the model equations, which are automatically rearranged into the forms required by these approaches. We have also integrated the CVODE solvers, a ‘gold standard’ for stiff systems, and we have developed support for symbolic computation of the Jacobian matrix, yielding further increases in the performance and accuracy of CVODE. We discuss some of the technical details of this work and compare the performance of the available numerical methods. Finally, we discuss how this is generalized in our functional curation framework, which uses a domain-specific language for defining complex experiments as a basis for comparison of model behavior.