In silico MS/MS spectra for identifying unknowns: a critical examination using CFM-ID algorithms and ENTACT mixture...

Area of Science:

• Analytical Chemistry
• Computational Chemistry
• Environmental Science

Background:

• High-resolution mass spectrometry (HRMS) is crucial for chemical analysis.
• Reference spectral libraries are limited by the availability of chemical standards.
• Expanding chemical identification capabilities is essential for complex mixture analysis.

Purpose of the Study:

• To evaluate the utility of in silico predicted MS/MS spectra for chemical identification.
• To assess the performance of in silico spectra against experimental data.
• To determine if in silico spectra can augment existing reference libraries.

Main Methods:

• Generated in silico MS/MS spectra for ~765,000 compounds using CFM-ID.
• Utilized experimental spectra from 10 EPA Non-Targeted Analysis Collaborative Trial (ENTACT) mixtures.
• Compared identification performance of in silico libraries versus commercial reference libraries.

Main Results:

• In silico libraries correctly identified up to 50% of 377 unique compounds from ENTACT mixtures.
• Commercial reference libraries identified approximately 53% of the compounds.
• The combined use of reference and in silico libraries correctly identified 73% of the compounds.
• In silico spectra demonstrated a true positive rate of 0.90 with variable false positive rates for candidate filtering.

Conclusions:

• In silico spectra effectively identify true positives in complex samples, comparable to reference spectra.
• In silico spectra serve as a valuable tool for filtering potential false positives.
• Augmenting reference libraries with in silico spectra significantly improves chemical identification rates.