Solubility-Weighted Index: fast and accurate prediction of protein solubility

Bikash K. Bhandari; Paul P. Gardner; Chun Shen Lim

To confirm the usefulness of SWI in solubility prediction, we compared it with the existing tools Protein-Sol (Hebditch et al. 2017 ) , CamSol v2.1 (Sormanni, Aprile, and Vendruscolo 2015; Sormanni et al. 2017) , PaRSnIP ) , DeepSol v0.3 (Khurana et al. 2018) , the Wilkinson-Harrison model (Davis et al. 1999; Harrison 2000; Wilkinson and Harrison 1991) , and ccSOL omics (Agostini et al. 2014 ) . We did not include the specialised tools that model protein structural information such as surface geometry, surface charges and solvent accessibility because these tools require prior knowledge of protein tertiary Fig 4A) . Our SWI C program is also the fastest solubility prediction algorithm (Table 1, Fig 4B and Supplementary Table S7 ). The wall time was reported at the level of machine precision (mean seconds ± standard deviation). A total of 10 sequences were chosen from the PSI:Biology and eSOL datasets, related to Fig 4B and Supplementary Table S7 (see Methods). b For SWI, mean AUC ± standard deviation was calculated from a 10-fold cross-validation (see Methods). For other tools, no cross-validations were done as the AUC scores were calculated directly from the individual subsets used for cross-validation. c DeepSol reports solubility prediction as probability and binary classes. The probability of solubility was used to calculate AUC and Spearman's correlation due to better results. AUC, Area Under the ROC Curve; NA, not applicable; PDB, Protein Data Bank; PSI:Biology, Protein Structure Initiative:Biology; ROC, Receiver Operating Characteristic; R s , Spearman's rho; SWI, Solubility-Weighted Index; s, seconds.