Phase separation is exploited by biology to organize a number of cellular processes. Examples include heterochromatin in transcription regulation, ribonucleoprotein granules for spatial control of translation in neurons, and phase separation as a key mechanism for the functional organization of the cell membrane. Phase transitions also play important roles in the development of human disease. For instance, RNA phase transition may be involved in severe neurological conditions such as Huntington disease and muscular dystrophy . Liquid-liquid phase separation promotes the aggregation of the intrinsically disordered protein tau to amyloid fibers, which may play a crucial role in the pathogenesis of Alzheimer’s disease . Cataract ensues from the formation of insoluble protein aggregates that affect the transparency of the eye lens .
Numerical tools designed to study biology at the molecular and cellular level have seen a dramatic development in the past decades and seem ripe to provide insight into the roles and mechanism of phase separation in health and disease. At the nanoscale, all-atom molecular dynamics have proved their usefulness as they can provide detailed insight into intimate microscopic mechanisms. However, their most critical weakness lies in the attainable timescales when considering systems with a size of biological interest. Simulation times are generally limited to a dozen of microseconds, but can be pushed up to a fraction of millisecond using large-scale high-performance computing facilities. However, timescales of interest in molecular and cell biology are far longer, between the millisecond and the second, and beyond. In this context, alternative methods have been developed to reach longer timescales.
A first step consists in considering small groups of atoms (e.g., one or few water molecules, a methyl group) instead of single atoms, which are seen as a single effective particle, thus leading to coarse-grained models. Another promising approach consists in going a step further in the coarse-graining process, thus giving up the ambition of accounting for realistic interactions at the atomic level. Effective mesoscopic models are then at stake, where elementary “particles” are as large as molecules (e.g., lipids, proteins) or subparts of macromolecules (e.g., bases or base-pairs in the case of nucleic acids, actin monomers). In any case, the relevant degrees of freedom must be identified, and then the model parameters must be finely tuned to account at best for the microscopic degrees of freedom that are neglected in the coarse-graining process. This can be achieved by using a high-resolution model to determine the parameters of a lower-resolution one. The gain in terms of simulation times and attainable system sizes is then of several orders of magnitude. When increasing the degree of coarse-graining, biologically relevant questions can be addressed in a wider class of cases and the resultant increase in physical insight can be substantial. Last but not least, analytical approaches using mesoscopic models from statistical physics, eventually coupled to hydrodynamics, elasticity theory, or electrostatics depending on the biological question, are always very useful to understand the fundamental mechanisms at play and to span a wider range of parameters in spite of inevitable approximations.
Modeling phase transitions leads to inherent challenges from a computational perspective. Critical slowing down near continuous phase transitions or slow coarsening processes in phase-separated systems require the development of specific approaches. Time-scales of interest go from the microsecond (or even shorter) at the molecular scale to hours or days when considering the whole cellular mechanism. Length scales of interest are much larger than atomic scales and using coarse-grained or mesoscopic models is in general mandatory. Bridging the different scales is one of the big challenges in this context, and integration of experimental information in computational and theoretical modelling is crucial.
The workshop will be focused on three main domains (without exhaustivity), as follows. Note that each cell subpart (membrane, cytoplasm, nucleus, etc.) requires specific modeling or experimental tools. However, in a cell these subparts strongly interact with each-other and mixing people working on these different entities will benefit to the whole audience.
a. Cell membrane and lipids
As a complex mixture of interacting biomolecules embedded in a lipid mattress, cell membranes are the focus of an extensive literature in which modeling takes an increasing part. Some popular force fields such as the MARTINI one have specifically been designed to study them . One of the central issues in these systems is to account from a theoretical perspective for the nanodomains (a paradigmatic instance of which are named “rafts”) commonly observed by modern microscopy techniques in biomembranes, and resulting from phase separation of the membrane constituents [5-9]. In this respect, much progress has been accomplished in the last decades, but many issues remain. Filling the gaps between the different scales (from molecules to cells) is notoriously challenging. Many common drugs target membrane proteins and many pathologies involve biological mechanisms occurring at or close to the plasma membrane, which increases further the interest of understanding the organization of these systems in depth. A large amount of biological functions, such as viral and bacterial infection, immune response, cell adhesion, transport of solutes or signaling, to name a few, are concerned.
b. Cytoplasm and proteins
The cell plasma membrane delimits the cellular medium, itself a rich and crowded, yet organized mixture of solutes, proteins, biopolymers and organelles. In prokaryotic cells, organelles are not required to organize the cytoplasmic functions, however a certain degree of organization prevails and phase separation is evoked as a leading mechanism to avoid random localization within the cytoplasm [10,11]. Phase separation of cellular material into macroscopic liquid-like droplets has recently emerged as a fundamental organizing principle in eukaryotic cells [12,13]. Here, three 3D phase separation gives rise to essential “membraneless organelles” such as the nucleolus  and germ granules . Understanding critical nucleus formation at the onset of phase separation in this context, and how the interactions of components at the surface of membraneless organelles differ from those in their interior have so far evaded experimentalists. Insightful simulations in these areas have the potential to propel the field forwards by generating new and testable hypotheses.
c. Nucleus, chromatin, and nucleic acids
One of the most studied organelles of mammalian cells is their nucleus, which is the hub for gene regulation. The genome is highly compacted to fit within the nucleus; however, it must be accessible to the transcriptional machinery to allow appropriate expression of genes. Understanding of how nuclear organization reconciles these two somewhat contradictory requirements is challenging. Embracing a hierarchy of scales, from the mechanical properties at the base-pair scale and epigenetic regulation, to chromatin structure and to the description of chromosome territories, the understanding of the functional genome has seen impressive progress in the recent years. However, our understanding of the link between nuclear organization and the whole process of genetic regulation, including its dysfunctions in genetic diseases and cancer, remains fragmented. Models making the connection between the different scales are absolutely needed to understand how this organization emerges from the interactions of constituent nuclear components. Notably, the dynamic spatial organization of genes is of primary importance for their mutual interactions (activation or silencing) , within and between chromosomes, but a unifying physical model of organization is still lacking [17,18].