Sudden cardiac arrest is the leading cause of death, stated Matthias Reumann. In the USA, there are about 424,000 deaths per year due to cardiac arrest. That is one casualty every 1 minute and 15 seconds. Globally, there are 6.3 million deaths per year which is one every 5 seconds.
The bad news is that drugs can promote arrhythmia and SCA, and can even cause anti-arrhytmics. These irregular and chaotic electrical disturbances in the heart cause a lot of reason to worry.
Matthias Reumann explained the anatomy and electrophysiology of arrythmia. The heart pumps blood through the body with electrical activity. The ions are moving across the cells.
There is a multi-scale reaction for the diffusion of cardiac models with non-linear stiff ODEs. The 3 to 100 state variables per cell are depending on the model's detail. We are talking about 370 million of tissue cells, stated Matthias Reumann.
Matthias Reumann started in 2007 to model electrophysiology in the heart. One single heartbeat would cost 2 weeks to simulate. Cardioid however is 1200 times faster than the state-of-the-art software, he told the audience with 58,8% of peak performance on the full Sequoia, which amounts to about 1,6 million cores.
Matthias Reumann expanded on the Big Data challenge that we are facing. When researchers are just looking at l/m - 8 bytes (double precision) with 0.05 mm resolution, 3 billion elements are involved, costing 1 petaByte per 1 minute simulation.
A single cell means 1 byte and 1 minute simulation. In basic science, we are talking 8x1x60,000, meaning 48 kiloByte per 1 minute simulation, stated Matthias Reumann.
For the experimental ventricular wedge preparation, methods involve the construction of a 1-D cable. The in silico model is an independent risk factor. Predictive modelling involves risk assessment.
Predictive modelling on drug effects is amounting to 8 x 200 x 200 x 200 x 60,000, calculated Matthias Reumann. Are those wedges predictive for the whole heart? In fact, the heart has very fine structures. It is not clear how the distribution of these structures is, explained Matthias Reumann.
For the predictive modelling of the whole heart, scientists require 0,05 mm resolution, meaning 1 petaByte.
A 0,1 mm resolution amounts to 0.1 Petabyte.
Matthias Reumann presented a number of open questions to the audience:
Matthias Reumann showed the example of a simulation of E-4031 and Ranolazine with the S1-S2 protocol.
The Genome Wide Association Studies (GWAS) are being used to identify common genetic factors that influence health and disease. Is a single gene related to a phenotypic trait? Researchers use an univariate analysis.
The bottleneck is the computation time, explained Matthias Reumann.
A study with 1 million SNPs has 500 billion SNP pairs or 1667 trillion SNP triplets. Increasingly large datasets cannot be stored in the memory.
Is a single gene related to a phenotypic trait? Gene-gene and gene-environment interactions are computationally expensive. There are about 10 million single nucleotide polymorphisms in the human genome.
The results include a two-way interaction of 2000 samples of 1.1 million SNPs, computed in about 12 minutes, stated Matthias Reumann.
What if the researchers find something new? The GWAS depth analysis is being performed. There are SNPs in about 40 candidate genes.
Computational, data driven analytics involves machine learning, interpretation, data extraction, exploration, and cataloguing.
The domain knowledge includes spatial modelling of cellular processes and the whole cell. The problem, according to Matthias Reumann, is which impact does the cell shape have? This is described in a paper titled "The dawn of Virtual Cell Biology".
Simulating the cell is performed with network dynamics showing differential equations on a simulation time scale.
The molecular dynamics simulation is based on first principles, stated Matthias Reumann. The time scale amounts to just micro seconds which is not sufficient.
Matthias Reumann concluded by saying that Big Data and analytics are enormous challenges in the life sciences. Researchers still have a lot of thinking to do.