“…3-D structurebased methods include the prediction of soluble expression by molecular dynamics (MD)-simulated unfolding combined with a support vector machine (SVM) architecture (Schaller et al, 2015), dynamic exposure of hydrophobic patches in MD simulations (Chennamsetty et al, 2009;Jamroz et al, 2014), and projection of sequence-based methods onto 3-D structures (Sormanni et al, 2015;Zambrano et al, 2015). Although high-throughput 3-D structure generation of VLP building blocks has been described previously (Klijn et al, 2019), the computational cost of creating 3-D structures is still high, limiting the applicability of this approach in candidate selection for several hundred molecules. Amino acid sequence-based methods can be distinguished into amino acid composition-based algorithms such as machine learning approaches using SVM or random forest classifiers (Magnan et al, 2009;Agostini et al, 2012;Samak et al, 2012;Xiaohui et al, 2014;Yang et al, 2016) and sliding-window-based algorithms, such as AGGRESCAN, Zyggregator, and CamSol (Conchillo-Sole et al, 2007;Tartaglia et al, 2008;Sormanni et al, 2015).…”