Hydrodynamics interactions (HIs) are known to heavily influence the
slow-down of the mobilities of colloids in crowded environment. For example,
we have recently used Brownian dynamic simulations to study the
diffusion of proteins in crowded environment and found that HIs are a major
contributor to the slow down of proteins under these conditions. From
a computational point of view, the inclusion of HIs in implicit solvent simulation
(as in Brownian Dynamics) can be highly costly in terms of computer
time. To address this limitation, a common strategy is to use a mean-field
approximation to include HIs. In this approach, HIs are modeled as corrections
to the diffusion coefficients of the colloids. For this, expressions of
how the diffusion coefficients depend on the occupied volume fraction are
used, but this approach does not completely account for the local distribution
of colloids. In this work, we introduce a data driven model that predicts
the diffusion coefficient of colloids and includes far field and lubrication HIs
[4]. We will also discuss how this approach can reduce the computational
cost compared to other methodologies (e.g. Stokesian Dynamics) and how
polydispersity (which is normally ignored) can be also addressed with this
methodology.