|Overview of dataset|
|No of total compounds||General 307 low molecular weight compounds|
|No of C atoms/compound||1–9|
|M.W.||32.0–252.7 (ave. 98.6)|
|Solubility data||HSPs (δD, δH, δP), SP & LogP|
|Analytical data||H/C-NMR assignments, density & refractive index|
|Molecular data||Molecular components (7 items) & RDKIT's descriptors (14 items)|
|Algorithm||in-phases (3-step) deep neural network|
|Validation method||5-division cross-validation|
|Hyper parameter selection||Bayesian optimization method|
|R2||δD:0.81, δH:0.61, δP:0.61, SP:0.58, LogP:0.69|
|RMSE||δD:0.51, δH:3.00, δP:2.16, SP:1.85, LogP:0.59|
MSC in BriefThe solubility, such as HSPs composed three elements (δD, δH, δP), SP, and LogP, is one of important physical parameter for understanding the characteristics of a substance. These solubility parameters have been used in several situations (e.g., materials science, biorefinery, cosmetic chemistry, and drug discovery). Here we've developed a useful and agile materials informatics tool, named MatSolCa (MSC), as solubility calculator created with special DNN model. The tool can calculate solubility parameters of HSPs, SP and LogP from only analytical data, such as 1D-H/C-NMR data, density and refractive index. We have expectations for effective R & D with MSC.
CitationAtsushi KUROTANI, Toshifumi KAKIUCHI and Jun KIKUCHI, Solubility prediction from molecular properties and analytical data using an in-phase deep neural network (ip-DNN). ACS Omega, 2021, vol6, 22, 14278-14287
PolicyWe just use your input data for solubility calculation.