The transition between the native and amyloid states of proteins can proceed via a deposition pathway via oligomeric intermediates or via a condensation pathway via liquid droplet intermediates generated through liquid-liquid phase separation. While several computational methods are available to perform sequence-based predictions of the propensity of proteins to aggregate via the deposition pathway, much less is known about the physico-chemical principles that underlie aggregation within condensates. Here we investigate the sequence determinants of aggregation via the condensation pathway, and identify three features, the droplet-promoting propensity, the aggregation-promoting propensity and the multimodal interactions quantified by the binding mode entropy. By using this approach, we predict aggregation-promoting mutations in droplet-forming proteins associated with amyotrophic lateral sclerosis (ALS). This analysis provides insights into the amino acid code for the conversion of proteins between liquid-like and solid-like condensates.

蛋白质的天然状态和淀粉样蛋白状态之间的转变可以通过低聚中间体的沉积途径或通过液-液相分离产生的液滴中间体的缩合途径进行。虽然有几种计算方法可用于对蛋白质通过沉积途径聚集的倾向进行基于序列的预测，但对凝聚物中聚集的物理化学原理知之甚少。在这里，我们通过凝聚途径研究聚集的序列决定因素，并确定了三个特征，即促进液滴的倾向、促进聚集的倾向和结合模式熵量化的多模式相互作用。通过使用这种方法，我们预测与肌萎缩侧索硬化症（ALS）相关的液滴形成蛋白中的聚集促进突变。该分析提供了对蛋白质在液体状和固体状冷凝物之间转化的氨基酸密码的见解。