About this Research Topic
Modelling of biological systems has traditionally been pursued along one of two distinct research directions: Bottom-up theory-driven modelling from first principles and top-down data-driven modelling. While theory-driven modelling has the advantages of having a direct link between the model terms and known mechanisms of the system, and being free from bias and noise, the models are often oversimplified, prone to numerical instabilities and difficult to parametrize based on measured data. It is also difficult to assimilate long-term historical data into the models, and the model output is very sensitive to uncertainties in the input, like boundary conditions, initial conditions and input parameters. Large mechanistic models for complex systems, such as biological systems, that account for the mechanisms on a detailed level, also become computationally demanding. This exaggerates the problems with parametrizing the models, since the models need to be run many times with different input settings in order to be fitted to measured data.
Data-driven modelling on the other hand, can handle large amounts of data collected from e.g. a biological system, and model the relationships between the variables without any knowledge of the underlying mechanisms. This field has lately experienced a rapid development of new methodology, and very complex systems can now be modelled accurately with multivariate data analysis, machine learning and deep learning. Data-driven models are stable for making predictions, and long-term historical data can be taken into account. However, they face challenges with overfitting, random correlations, bias, noisy data, as well as difficulties in model interpretation (black-box).
Hence, in order to fully utilise the potential that lies in the large amounts of data that are being collected from e.g. imaging, -omics, sensors and spectroscopy, we need robust data analysis methods that are suitable for modelling complex, nonlinear problems in a wide range of applications, but to be able to interpret what we see from the data and avoid false discovery, we also need a link to prior knowledge of the system and the underlying mechanisms. A combination of the two modelling directions therefore has great potential in facilitating new discovery.
In this Research Topic, we welcome manuscripts that lay in the intersection between these two modelling traditions.
We welcome contributions in any form, both theoretical and applied to biological problems from diverse fields, including but not limited to:
- Dynamic modelling of biological systems
- Model parameterization
- Multi-scale modelling
- Interpretable AI
- Inverse modelling
- Sensitivity analysis
Topic Editor Hugo Geerts is employed by Centara, US. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
Data-driven modelling on the other hand, can handle large amounts of data collected from e.g. a biological system, and model the relationships between the variables without any knowledge of the underlying mechanisms. This field has lately experienced a rapid development of new methodology, and very complex systems can now be modelled accurately with multivariate data analysis, machine learning and deep learning. Data-driven models are stable for making predictions, and long-term historical data can be taken into account. However, they face challenges with overfitting, random correlations, bias, noisy data, as well as difficulties in model interpretation (black-box).
Hence, in order to fully utilise the potential that lies in the large amounts of data that are being collected from e.g. imaging, -omics, sensors and spectroscopy, we need robust data analysis methods that are suitable for modelling complex, nonlinear problems in a wide range of applications, but to be able to interpret what we see from the data and avoid false discovery, we also need a link to prior knowledge of the system and the underlying mechanisms. A combination of the two modelling directions therefore has great potential in facilitating new discovery.
In this Research Topic, we welcome manuscripts that lay in the intersection between these two modelling traditions.
We welcome contributions in any form, both theoretical and applied to biological problems from diverse fields, including but not limited to:
- Dynamic modelling of biological systems
- Model parameterization
- Multi-scale modelling
- Interpretable AI
- Inverse modelling
- Sensitivity analysis
Topic Editor Hugo Geerts is employed by Centara, US. All other Topic Editors declare no competing interests with regards to the Research Topic subject.
Keywords: Mechanistic modelling, data-driven modelling, bottom-up, top-down, dynamic modelling, multi-scale modelling, AI
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
