Red Blood Cell Parameter Inference (Relaxation)

An interesting aspect of RBCs that have been observed experimentally is the effect that viscosity has on relaxation time of an RBC membrane. This phenomenon is, however, still not fully understood from a computational point of view. To gain a deeper understanding of the viscosity pa-rameter on RBC computational models, we studied the pairwise dissipation interaction between neighboring ver-tices on the triangular mesh. One of the unambiguous experiments to determine the membrane viscosity is the relaxation of a stretched RBC to its equilibrium shape. Due to the presence of heterogeneous data from five dif-ferent experimental studies, we formulated the inference problem for the parameter governing the dissipation in-teraction as a two-staged hierarchical Bayesian inference to estimate the membrane dissipation and its uncertainty, as well as reason about the validity of the RBC model implemented in Mirheo.

In the first stage, we inferred five posterior distributions for the dissipation parameter, each conditioned on an individual data set. To infer these distributions, we defined a likelihood function comprising the execution of a virtual RBC relaxation experiment in Mirheo that allows us to compare simulated length scales to experimental measurements. In the second stage, we assumed that all the posterior distributions found in the first stage follow a generalized distribution that is controlled by some hyperparameters which remained to be inferred.

We performed the sampling in stage one and two using BASIS implemented in Korali, whereas we used 512 samples in stage one, and 10000 samples in stage two. Finally we found that the estimated membrane dissipation parameter, by the means of the maximum a posteriori (MAP), corresponds to membrane viscosity parameters estimated in literature.

In this case study, we sample five posterior distribu-tions of the membrane dissipation parameter (γC) during stage one. All five experiments share a similar setup and differ only in the initial conditions and the reference data.

Scientific Sources

Data in set 1 in [data/hochmuth_1979](data/hochmuth_1979):

Hochmuth, R. M., Worthy, P. R., & Evans, E. A. (1979). Red cell extensional recovery and the determination of membrane viscosity. Biophysical Journal, 26(1), 101–114. https://doi.org/10.1016/S0006-3495(79)85238-8

Data set 2 in [data/henon_1999](data/henon_1999):

Hénon, S., Lenormand, G., Richert, A., & Gallet, F. (1999). A new determination of the shear modulus of the human erythrocyte membrane using optical tweezers. Biophysical Journal, 76(2), 1145–1151. https://doi.org/10.1016/S0006-3495(99)77279-6

This experiment is based on [this](https://github.com/cselab/mirRun/tree/master/rbc_relaxation/) setup.

Results Report

  1. Wälchli, S. Martin, A. Economides, L. Amoudruz, G. Arampatzis, P. Koumoutsakos, Load Balancing in Large Scale Bayesian Inference, in: Proceedings of the The Platform for Advanced Scientific Computing (PASC) Conference 2020, to appear, 2020

Setup

Installation on Piz Daint

Install Mirheo Tools

cd $HOME
mkdir src
cd src
git clone --recursive https://github.com/cselab/Mirheo.git
cd Mirheo
cd tools
./configure --bin-prefix $HOME/.local/bin --exec-cmd srun
make
make install

Install Mirheo

cd $HOME/src/Mirheo
. mir.load
mir.make install
cd ..
rm -rf Mirheo

Install python modulddes

python -m pip install trimesh --user
python -m pip install mpi4py --user

Running

mir.run --runargs "-n $SLURM_JOB_NUM_NODES --ntasks-per-node=1" ./run.py