Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry

Tobias Kind, Oliver FiehnBMC Bioinformatics 2007 · DOI

Introduction

In analytical chemistry, identifying an unknown molecule typically begins with mass spectrometry (MS). Modern high-resolution instruments provide "accurate mass," measuring a molecule's weight to several decimal places. However, mass alone is rarely sufficient for identification. As molecular weight increases—particularly above 500 Da—a single mass measurement can correspond to hundreds or even thousands of theoretical elemental compositions. This phenomenon, known as "formula explosion," presents a significant challenge: how to distinguish chemically valid formulas from mathematically possible but biologically impossible permutations.

This paper introduces the "Seven Golden Rules," a set of heuristic and chemical filters designed to apply "chemical common sense" to mass spectrometry data. By filtering out noise and ranking candidates, these rules achieve a 98% success rate for compounds present in databases and significantly narrow the search for novel molecules.

The Seven Golden Rules

The algorithm utilizes seven distinct filters derived from the analysis of millions of known compounds across repositories like PubChem, Wiley, and the Dictionary of Natural Products (DNP).

  1. Restrictions on Element Numbers: Setting upper limits for specific atoms based on total mass (e.g., preventing unrealistic nitrogen counts in small molecules).
  2. LEWIS and SENIOR Rules: Deterministic rules regarding valency and the octet rule to ensure a formula can form stable chemical bonds.
  3. Isotopic Pattern Filter: Comparing theoretical isotope abundances (e.g., $^{13}C$, $^{34}S$, or halogens) against measured spectral data. This is often more informative than mass accuracy alone.
M+1 and M+2 isotopic abundances distinguish between different chemical classes.
  1. Hydrogen/Carbon Ratios: Filtering based on the observation that most organic molecules fall within a specific H/C ratio range (typically 0.1 to 6).
Histogram of H/C ratios showing the constrained chemical space of known molecules.
  1. Heteroatom Ratios: Checking ratios of N, O, P, and S relative to Carbon (e.g., O/C ratios rarely exceed 1.2).
  2. Element Ratio Probabilities: A sophisticated check for molecules containing multiple heteroatoms (NOPS) to ensure the combination is statistically likely.
  3. TMS Check: A specialized rule for GC-MS data to account for trimethylsilyl derivatives used during sample preparation.

Key Insights: Isotopes and Nature's Constraints

A primary finding of the study is that isotopes are often more valuable than mass accuracy. Even with sub-1 ppm mass accuracy, unique identification is frequently impossible without isotope information. The "M+1" and "M+2" peaks act as a secondary fingerprint, allowing researchers to identify specific elements like Bromine or Chlorine that have distinct isotopic signatures.

Furthermore, the authors demonstrate that "Nature is conservative." By analyzing millions of compounds, they found that chemical space is surprisingly constrained. Most molecules follow specific "ratio corridors," allowing the Seven Golden Rules to exclude over 90% of mathematically possible formulas. For formulas up to 2000 Da, the rules reduced the search space from 8 billion theoretical compositions to just 623 million.

Results and Performance

The rules were validated against 432,968 unique formulas in PubChem. Only 0.6% of known compounds failed the filters, proving they are robust without being overly restrictive. When integrated with database queries, the correct formula was retrieved as the top hit in 80–99% of cases. For truly novel compounds not found in any database, the correct formula still appeared within the top three candidates 65–81% of the time.

Limitations

As these rules are heuristic, they may fail for highly unusual synthetic compounds or extreme molecules (such as methane or highly halogenated substances) that fall outside common ranges. Additionally, the rules depend on the correct identification of the "adduct" (the ion picked up during ionization) and the quality of the instrument's isotope intensity measurements. Future work will focus on automated adduct recognition and the integration of fragmentation data (MS/MS) to further refine isomer ranking.

Summary updated

BibTeX
@article{kind2007seven,
  author = {Tobias Kind and Oliver Fiehn},
  title = {{Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry}},
  journal = {BMC Bioinformatics},
  year = {2007},
  doi = {10.1186/1471-2105-8-105},
}
Figure