Unsupervised Pharmaceutical Polymorph Identification and Multicomponent Particle Mapping of ToF-SIMS Data by...

Area of Science:

• Analytical Chemistry
• Materials Science
• Computational Chemistry

Background:

• Pharmaceutical crystal polymorphism significantly influences physicochemical properties and API production.
• Characterizing polymorphs at the single-particle level is crucial for targeted manufacturing.
• Traditional analysis of high-resolution chemical imaging data, like ToF-SIMS, can be laborious and prone to user bias.

Purpose of the Study:

• To apply unsupervised machine learning, specifically NMFk, for analyzing ToF-SIMS chemical imaging data of inkjet-printed acetaminophen.
• To identify and chemically map different polymorph phases of acetaminophen.
• To demonstrate NMFk's capability in deconvoluting mixed pharmaceutical samples.

Main Methods:

• Time-of-flight secondary ion mass spectrometry (ToF-SIMS) for chemical imaging.
• Non-negative matrix factorization (NMF) and its variant NMFk for unsupervised data analysis.
• K-means clustering integrated within NMFk for dimensionality determination and optimization.

Main Results:

• NMFk successfully identified three polymorph phases of acetaminophen: amorphous, crystalline form I, and crystalline form II.
• The method generated representative mass spectra for each polymorph and mapped them onto particle samples.
• NMFk effectively decomposed mixed pharmaceutical samples, identifying constituent compounds and their compositions.

Conclusions:

• NMFk is a powerful unsupervised tool for analyzing complex ToF-SIMS data, enabling polymorph identification and mapping.
• This approach advances the chemical characterization of pharmaceutical materials at the single-particle scale.
• NMFk facilitates the analysis of multi-component pharmaceutical formulations and aids in compound identification via spectral library matching.