Experimental Models

There have been ground-breaking developments in model systems and biotechnologies for dementia research. This means that we now have more experimental models than ever, including a spectrum of mice with specific genetic mutations or knock-ins, patient-derived iPSC derived cell cultures (1) and improved sampling of human tissues (2). We have additionally seen the development of complex multi-cellular and multi-species models, including organoids of the human brain (3), multi-species models of specific the blood-brain barrier (4) and chimeric mouse models, containing live human cells (5). All these models capture different aspects and states of disease biology and allow varying extents of control over genetic and environmental experimental intervention. Together they offer increasing scope and potential to elucidate the biological mechanisms directly involved in neurodegenerative phenotypes and diseases. However, this leaves open the question as to how suitable any given model is to understand specific aspects of human neurodegeneration, and what aspects of human disease it can faithfully replicate. Secondly, work on experimental model systems is frequently conducted in isolation with (often small-scale) single studies working on an experimental model and asking one constrained question. Employing the data science expertise of the DEMON network, we will lay out a vision for a data science driven experimental medicine and tackle the challenge of integrative analyses across multiple studies and heterogeneous model systems.

Specifically, the experimental models working group aims to progress the field of experimental medicine in dementia research by addressing the following big questions:

1. What makes a good experimental model?

The question of how to quantify validity in modern experimental medicine is of core importance to the experimental models working group. Beyond descriptive face validity we aim to discuss and work out measurable criteria for how to determine suitability and representativeness of experimental models. Which aspects of disease biology can be captured by various model systems and, in turn, what questions can be answered by a given model system. These answers are straightforward in some cases: for example, neuron-glia interactions cannot be modelled in a neuron-only cell culture. However, the answer can be more subtle and complex in other cases: for example, given that iPSC-derived cell cultures or organoids show gene regulatory ageing markers of very early development, do they represent suitable models of neurodegenerative disease processes in the context of old age? Or could ageing signatures even be simulated experimentally or computationally in these in vitro systems?

• How can we make experimental medicine more reproducible?

By and large experimental medicine and neuroscience in dementia research has relied on experiments conducted in small samples. This directly causes problems with statistical power as well as an inflation of false positive associations. Compounded additionally by publication bias it is difficult to determine what the true replication rate of such studies is. Furthermore, robust studies should control for experimental factors and batch effects that can have an impact on the measured phenotype, by harmonising, randomising across experimental conditions and including appropriate covariates in the statistical analysis. This comes at a cost of statistical power and requires large sample sizes, as the degrees of freedom are reduced by adding additional covariates. However, it is crucial that these factors are measured and controlled for where possible, as they have been shown to lead to dramatically different outcomes across experiments. This has been illustrated, for example, in a recent multi-centre study of iPSC derived neurons, with substantial batch effects across centres, despite harmonised experimental protocols (6). To better understand sources of variation and how to handle them experimentally and statistically these types of studies will be crucial going forward. As a first step towards greater replication of scientific results, we also advocate for better reproducibility of studies. Reproducibility across reporting of experimental protocols, code and analyses as well as making experimental data available facilitates verification of given results and more importantly increases the utility of the study to the field, given that it can be used to answer further questions. Open reporting and data availability will also facilitate meta- and mega-analytic approaches leveraging the raw data, leading to more robust results in the field.

• Multi modal insights into dementia research?

One of the key directions for dementia research and experimental models specifically will be generating layered multi-modal datasets. Being able to connect brain activity with gene expression patterns, for example could give entirely new insights into gene regulation in the context of functional activation of neurons. In the same way, going forward atlases across phenotypes and omics, carefully collected on matched samples are set to provide novel insights into disease biology. From the standpoint of the DEMON network, the development analysis methods and tools that go across modalities and leverage these links, is of great interest. This will become a rapidly evolving field and one where the application of machine learning and AI approaches is extremely promising. 

• How can we translate insights from experimental models to human disease biology?

A key question for the experimental models working group is how insights from experimental systems can be translated more intelligently to human disease biology. It is clear that experimental medicine and drug discovery does not always translate directly into human application. The number of failed clinical trials that result from drugs developed in animal models is unfortunately growing. Beyond general considerations of how to make research more translatable, we will actively work on quantitative models for cross-model translation. We are particularly interested in leveraging ML approaches to translate gene-regulatory networks and the response to experimental perturbation across species into human. We are also actively reviewing existing approaches (7) and those that could be adapted from unrelated field to address these kinds of translational challenges. Based on prior biological knowledge and large-scale reference datasets for baseline and perturbed conditions, we will develop novel translation algorithms.

1. Penney, Jay, William T. Ralvenius, and Li-Huei Tsai. “Modeling Alzheimer’s disease with iPSC-derived brain cells.” Molecular psychiatry 25.1 (2020): 148-167.
2. Nott, Alexi, et al. “Nuclei isolation of multiple brain cell types for omics interrogation.” Nature protocols (2021): 1-18.
3. Grenier, Karl, Jennifer Kao, and Phedias Diamandis. “Three-dimensional modeling of human neurodegeneration: brain organoids coming of age.” Molecular psychiatry 25.2 (2020): 254-274.
4. Gerhartl, Anna, et al. “The pivotal role of micro-environmental cells in a human blood–brain barrier in vitro model of cerebral ischemia: functional and transcriptomic analysis.” Fluids and Barriers of the CNS 17.1 (2020): 1-17.
5. Mancuso, Renzo, et al. “Stem-cell-derived human microglia transplanted in mouse brain to study human disease.” Nature neuroscience 22.12 (2019): 2111-2116.
6. Volpato, Viola, et al. “Reproducibility of molecular phenotypes after long-term differentiation to human iPSC-derived neurons: a multi-site omics study.” Stem cell reports 11.4 (2018): 897-911.
7. Brubaker, Douglas K., and Douglas A. Lauffenburger. “Translating preclinical models to humans.” Science 367.6479 (2020): 742-743.