Wavelet Approximation of GRID Fields: Application to Quantitative Structure-Activity Relationships

Area of Science:

• Computational chemistry
• Cheminformatics
• Molecular modelling

Background:

• Molecular interaction fields, like those from the GRID program, are crucial for drug discovery applications such as virtual screening and 3D-QSAR.
• These fields represent favorable interaction sites on molecules but contain a vast number of data points, posing computational challenges and potential issues like variable correlation in 3D-QSAR.

Purpose of the Study:

• To introduce and validate the use of wavelet methods for approximating large molecular interaction field datasets.
• To demonstrate that wavelet approximation can significantly reduce data dimensionality while retaining critical information content.

Main Methods:

• Application of wavelet transform techniques to compress molecular interaction field data generated by the GRID program.
• Validation of the compressed data through its application in 3D-QSAR modelling and comparison with original GRID fields.

Main Results:

• Wavelet approximation successfully reduced the number of variables in GRID fields, achieving high levels of data compression.
• The approximated GRID fields, when used in 3D-QSAR, yielded comparable results to those obtained with the original, uncompressed fields.
• Significant reduction in computational requirements was observed when using the wavelet-approximated data.

Conclusions:

• Wavelet methods offer an efficient approach to handle large molecular interaction field datasets.
• This data compression technique preserves the predictive power of GRID fields for applications like 3D-QSAR.
• Wavelet approximation enhances computational efficiency in molecular modelling and virtual screening workflows.