Coupling CNNs with high-throughput microfluidics to efficiently probe biomolecular phase space
Traditional pipetting experiments are great. They allow us to test out lots of different conditions for a given protein, and then we can see under what conditions certain phases are observed. Unfortunately, proteins are expensive and really hard to get in large quantities; as a physical chemist, the last thing I want to be doing (and likely failing to do) is expressing protein. So how can we explore the phase space of proteins whilst being conservative of sample?
Microfluidics allows us to ‘pipette’ conditions on an extremely small scale, allowing us to save lots of protein sample. This means more data can be collected in phase space, or more experiments can be carried out after. Arter et. al. designed the first version of the setup. By flowing aqueous solutions perpendicular to an oil flow, picolitre water-in-oil droplets can be generated at a rapid rate (>1000/min). By varying the flow rates of input solutions, we can force each droplet to contain a slightly different condition. This allows us to generate a phase diagram for a protein with minimal sample (~20μL).
Microfluidic droplets containing biomolecules under different environmental conditions. The images correspond to simultaneous imaging at two different wavelengths corresponding to two different components.
References
2023
Linking modulation of bio-molecular phase behaviour with collective interactions
Daoyuan Qian, Hannes Ausserwoger, William E. Arter, Rob M. Scrutton, and 9 more authors
Bio-molecular condensates formed in the cytoplasm of cells are increasingly recognised as key spatiotemporal organisers of living matter and are implicated in a wide range of functional or pathological processes. This opens up a new avenue for condensate-based applications and a crucial step in controlling this process is to understand the underlying interactions driving their formation or dissolution. However, these condensates are highly multi-component assemblies and many inter-component interactions are present, rendering it difficult to identify key drivers of phase separation. In this work, we employ the recently formulated dominance analysis to modulations of condensate formation, centred around dilute phase solute concentration measurements. We posit that mechanisms of action of modulators can be categorised into 4 generic classes with respect to a target solute, motivated by theoretical insights. These classes serve as a general guide towards deducing possible mechanisms on the molecular level, which can be complemented by orthogonal measurements. As a case study, we investigate the modulation of suramin on condensates formed by G3BP1 and RNA, and the dominance measurements point towards a dissolution mechanism where suramin acts on G3BP1 to disrupt G3BP1/RNA interactions, as confirmed by a diffusional sizing assay. Our approach and the dominance framework have a high degree of adaptability and can be applied in many other condensate-forming systems.
