QSPR.pxp
is a packed WaveMetrics Igor Pro 8 experiment file containing macros and compiled functions to perform:
- Regularized (multivariate) multiple linear regression ((M)MLR)
- Univariate logistic regression (LGR)
These regressions use molecular descriptors as explanatory variables, along with other numerical parameters.
- Molecular descriptors can be imported from an HTF5 file generated by the program
chemical_descriptor.py
. - An example input file is
des.h5
. - Example input CSV files for MLR and LGR regressions are also provided.
- The program requires the HDF5-64.xop extension to read HDF5 descriptor files (
*.h5
). - On first use, you will be prompted to specify the folder where the procedure (
*.ipf
) files are located.
file_io.ipf
andservice.ipf
: service functionsgen_discrete.ipf
: Genetic algorithm routinesqspr.ipf
: Main procedure file containing MMLR and LGR regressionqspr_cross_validation.ipf
: Cross-validation using random data splittingLinReg.ipf
: Igor implementation of linear algebra for regularized linear and logistic regression
- Macros and procedures are available via the Macros and QSPR drop-down menus at the top of the Igor Pro window.
- Key macros:
MLR_Genetics
LGR_Genetics
(found under Macros)
- Important QSPR functions:
QSPR -> Descriptors -> CreateSet_HDF5_des
: Reads thedes.h5
molecular descriptors fileQSPR -> CVS import -> LoadCSV_table
: Reads input CSV files with dependent variables and explanatory parameters
- Set the target number of descriptors.
- The program uses genetic algorithm optimization to find a descriptor set minimizing RMS error (or logistic analog).
- Hashing and starring (under QSPR -> Hash and Star) exclude or require specific descriptors.
- Visualization options include plots, tables, and MP4 videos.
QSPR -> Sets
provides macros to save, export, and import data sets and fits.
For questions on how to use this software, please contact:
Eli Shkrob
📧 [email protected]
📧 [email protected]