On April 11-12, we had a new bi-yearly meeting in STRATIF-AI. This meeting was held in beautiful Nottwil, Switzerland, where the Swiss Paraplegic Research center (SPF) is situated. SPF is responsible for the work with policy and actions toward stakeholders to achieve ultimate clinical implementation, and they were also the hosts of this event.
We have these events every 6 months, and they serve both the purpose of team building and increased understanding of each others’ perspectives and serve to help us focus on the most urgent and timely topics. This time we focused primarily on the design of the 6 clinical studies, which ethical applications were to be submitted at the end of that month. Four of these studies are dedicated to the collection of patient data, aimed at training both the machine learning algorithms and the mechanistic aspects of digital twins. The remaining two studies focus on the real-world testing of eHealth apps within clinical settings. The largest of those will test whether the digital twin improves clinical health conversations, in 300 patients, compared with 300 matched controls.
Overall, the meeting was a success, and we have now passed the planning phase of the project, with requirement specifications and ethical plans, and are now moving into action: into prototype development of the apps, and towards the first pilot studies.
On Friday, Gunnar Cedersund will give a lecture in the NBIS AI seminar series. NBIS is the National Bioinformatics Infrastructure Sweden, which is a core-facility spread out over all universities in Sweden. In other words, at each university in Sweden, there are some local bioinformaticians located, which helps out in projects. NBIS also arranges a seminar series on artificial intelligence (AI), and the next such lecture – on Friday May 17, at 10AM CET – features Gunnar Cedersund. Gunnar will talk about our digital twins, and about how they can be used to integrate a variety of data, both classical bioinformatics multi-omics data and behavioral and imaging data, into a personal data vault, and into a personalized copy of a patients. More information, abstract, and link to the online lecture is available here: https://scilifelab.atlassian.net/wiki/spaces/IJC/pages/2780758082/Meeting+59+2024-05-17
Our EU project STRATIF-AI (https://stratif-ai.eu/en), in which we develop digital twins for prevention, acute treatment and rehabilitation of stroke, was presented as part of “Clinical applications of computer modeling” at Latin American Summer School for Computational Neuroscience (LASCON) in Sao Paulo Brazil Jan 2024.
Figure shows how modern medicine has been changing and indicates role of modeling in modern (future) clinical practice — part of personalized medicine, data-mining, self-monitoring, omics, surgical planning, medical education, research, screening, etc
A central part of our digital twins, which underlies all apps in the STRATIF-AI project, are the models we have for the brain. So far these models have mostly focused on the neurovascular coupling (NVC). i.e. on the vascular response to neuronal activity. In our latest Ph.D. thesis “Mathematical Modelling of Cerebral Metabolism: From Ion Channels to Metabolic Fluxes”, which was defended by Dr Nicolas Sundqvist last Friday, we have taken these models to a higher level: to also include more detailed cerebral metabolism.
In the first paper, we use MRS data from humans to develop a small model for central metabolism, which is linked to our NVC models. In the second paper, we focus on cell-specific contributions, and show that inter-neurons might be the most dominant cell type in the NVC response; this puts much of the existing fMRI data into question. In the third paper, we present a detailed model for the most metabolically expensive process in neurons: the ion channels. Finally, in the fourth paper, we have developed a method for how to develop even more detailed intracellular flux models, using 13C-labelled media and a new approach to validation of intracellular flux models.
Opponent was Dr João Duarte from Lund, and the examination committee consisted of Jeanette Hellgren Kotaleski (KTH), Arvind Kumar (KTH), and Katarina Nöh (Jülich Forschungscentrum).
During Feb 14 and 15, 2024, we at ISBgroup at Linköping University (LIU) had a few visitors, who joined us for an interactive 2-day workshop on UX-design for our stroke Prevention App development. The Prevention app is developed jointly between Linköping University (LiU) and Z2, as part of the STRATIF-AI Horizon-Europe project (stratif-ai.eu), and it is meant for both primary and secondary stroke prevention. Primary prevention will be done in Region Västerbotten and Region Östergötland, and it involves a so-called Health Dialogue. A Health Dialogue is preceded by measurements of traditional risk factors (blood pressure, blood lipids, fasting glucose, etc), which are discussed together with e.g. a specialized nurse or a physiotherapist. Our app will use digital twins to help explain why a person should care about these risk factors, and how they can change their lifestyle (e.g. exercise more, eat less, take certain medications, etc), to improve the risk factors. Based on this information, the patient then sets some goals, to improve their lifestyle, to reduce one or several of the risk factors. These goals are then worked further on together with a coach and dedicated support-group. Also this subsequent coaching-period will be supported by the app, which thus provides a bridge between traditional healthcare actors, and those not normally integrated in healthcare: coaches, personal trainers, support groups. etc.
During the 2 days workshop, we did various role playing exercises, whiteboard and post-it note sketches, which then will be converted to Figma implementations, which then lead to new user tests, etc. From Z2, we have Jesper Fellenius, who has developed the backend for the personal data vault, where the data are stored, and who also has developed a prior prototype for a similar app, which we re-use. From LIU, we have Dr Gunnar Cedersund, Dr Dirk de Weerd, and M.Sc. student Greta Nilsson, who work with the app development. From Region Östergötland (RÖ), we have MD student Johanna Levander and MD specialist Valentin Kindesjö, who represent clinical needs. Apart from these, we also have Johan Holmsäter and Mats Janson, who represent Lev Skönare and coaching and wellbeing partners, who also are connected to STRATIF-AI.
This year, we got a double jackpot from the Swedish Research Council – who gave us glowing reviews for the 3R project, scaling from microphysiological in vitro systems to humans using scalable digital twins.
In Sweden, the Swedish Research Council (Swe: Vetenskapsrådet, VR), is the most central research grant, and it is often considered a key quality stamp of a top researcher to have at least one grant from VR. Therefore, competition is usually fierce (acceptance rate usually is 5-15%), and it is not at all guaranteed that you get money, even if you have a competitive application. Therefore, I am proud to say that this year, I got not only one grant, but two – and that the evaluation from the reviewers was unusually high and glowing.
The project I have gotten the review responses for so far is a special call on 3R, i.e. Replacement, Reduction and Refinement of animal experiments. This is a topic, I have been very active in ever since 2015, when I was awarded the first edition of the prize “Nytänkaren” (the thinker of new ideas), by the Swedish Fund for Research without Animal Experiments. The project extends on our experimental work, both within the group, doing cell biology cultures using 13C-labelled metabolites on liver and adipose tissue taken from surgery), and that in collaboration with AstraZeneca, centered around organs-on-a-chip, i.e. small microphysiological systems (MPS), with organoids and spheroids consisting of human cells (Fig A, recent paper). In the project, we will i) analyse these in vitro data using mechanistic modelling to get more information out of the data (e.g. metabolic fluxes), ii) plan new experiments, by first doing the experiments in the computer, and iii) translate the results to humans, by e.g. scaling the volumes of the spheroids to human sizes, and by adding the missing organs, which allows us to re-assemble the digital twin in the computer (Fig B-C, Step 1 and 2). The project will evaluate and quantify the benefits of this for e.g. drug development, and we will disseminate the results to pharma, scientists, and regulatory agencies (Fig C, Step 3).
In the evaluation, we only got 6s and 7s, which means that we were among the highest rated of all applicants, even among the few who got money (6 out of 56). The ranking is from 1-7, where a “normal, decent” scientist usually get a 3 (meaning “good”), and where you need at least a mixture of 5s and 6s to have any chance of getting money. If you get all 6s, you are usually getting the money for sure, and 7 is only very rarely given out (I was a reviewer for ~60 applicants two years ago, and then I think only one or possibly two got a 7 on any criterion). Therefore, I am very grateful that this year, I got only 6s and higher, and that two(!) categories got a 7: “merits of applicant” and “relevance for 3R” (Fig D). If the rating levels were the same as when I was a reviewer, I would – I think – have been number one of all applicants that year, and in any case, I must have been among the very top of all the 56 applicants also this year.
The life of a scientist is filled with many many rejected applications, so when you get a jackpot once in a while, it is important to stop a bit – and celebrate! Because tomorrow, it is time to get started working on the new exciting research projects! 🙂
Last week, I had the honor of giving the concluding keynote lecture at the event “Data-driven mechanistic modelling in life sciences”. This follows a trend of being invited to give more and more keynote and plenary lectures at events, for which I am very grateful. Such longer lectures also give me the chance to expand a little bit more on my point-of-view. The focus of this particular workshop is also something I am very keen to promote, since I think that this particular overlap (mechanistic and data-driven modelling) is under-represented in many communities and conferences.
The centrality of this overlap is actually seen even in our group logo (Fig A). a) The fact that it has an open non-black box, represents the fact that we do mechanistic modelling. b) The data-driven aspect of our models is represent by the purple core in the middle, which represents the fact that we always look for core predictions. Core predictions are predictions that are well-determined from the current prior knowledge and data, even when taking all uncertainties in data and prior knowledge into account.
While I personally think that this is the way to work, and while we have a very well-established workflow for how to develop models in this fashion, mechanistic modelling and data-driven modelling are unfortunately often done in two disjoint communities, with too little overlap (Fig B). Mechanistic modelling often results in mere simulation-based results, which have not been validated using independent data, i.e. data that has not been used to train the model. This is often the case for e.g. PDE and agent-based models, but also common in e.g. theoretical ecology, theoretical biology, etc. It was therefore encouraging too see that one of the presentations at the workshop (by Joshua Bull) looked at spatial models, and on how to quantify the comparison between simulations and data also for spatial models. Data-driven modelling is too often interpreted to mean only machine-learning, narrow AI, and other black-box modelling techniques. While these are big and very hyped communities and approaches at the moment, they are not the only techniques that can be used to do data-driven modelling. In other words, while these black-box models include important techniques, which are useful if one has standardized large-scale data, they also have critical short-comings. Black-box models e.g. have big problems incorporating the type of data that is present in most biological papers, including the prior that is knowledge available. For these reasons, explainability and trustworthiness are challenges. I therefore think that hybrid modelling is the way forward (see e.g. this review, and this example). At the conference, there was also an excellent opening keynote of day 2, by Alvaro Köhn-Luque, which showed some additional and interesting examples of hybrid models.
