COSMOquick 1.5 Previous released
COSMOquick 1.5 is available for Windows 64bit, Linux 64bit and MacOS. Please note: 32-bit operating systems are not supported any more.
New features include:
- Further extension of the .cosmo file database: The underlying database which allows for instant σ-profile generation generation has been further enriched. It contains now more than 190,000 .cosmo files and allows the fragmentation of most organic compounds.
- Global optimizer for solvent mixtures: The idea behind the solvent optimizer module is to find the best solvent (mixture) for the solutes, e.g. API and coformer simultaneously. The optimizer uses several boundary conditions such as avoiding solvates formation and assuring solvent miscibility. The optimizer is customizable and it can as well be used for single solutes or to find solvents that allow for the separation of two solutes.
- Improved machine-learning based QSPR models: The decision tree based model for the prediction of melting points has been further improved, it is built on a set of more than 18000 experimental data points and has a test set accuracy 37.2K (RSMD, root mean squared deviation). Besides, an additional model for the prediction of crystal densities based on more than 30000 experimental data points from the crystallography open database (COD, http://www.crystallography.net/cod/) is now available.
- Additional possibilities for high-throughput molecular descriptor generation: A large number of molecular descriptors (topological, SMARTS based, fingerprints) can now be generated very fast in memory and subsequently saved to .csv or .xls format for further processing.
- Improved batch conversion of molecule/compound names to molecular structures (2D/3D): the process of quickly generating 2D/3D structural information for a large set of molecules by starting just from a simple list of names has been improved. In particular the conversion of molecule/compound names into SMILES was improved by providing a large database. Alternatively batch requests can be sent to external web-services like PubChem (https://pubchem.ncbi.nlm.nih.gov/) to resolve the SMILES. The subsequent automatic 3D structure generation can be carried out via code based on the RDKit (http://www.rdkit.org/).
- Creation of high quality .sdf files: additional support for the conversion of 3D coordinates (.xyz or .cosmo files) into .sdf files containing high quality bond information was implemented. The approach uses a sophisticated bond parsing algorithm of the underlying COSMOfrag code.
- Various fixes and other improvements.