All About the New Annotation Confidence Scoring Node in Compound Discoverer Software

Anyone working in untargeted small molecule analysis knows that finding candidate annotations is only part of the job. The harder part is explaining how much confidence to place in each assignment, especially when the evidence comes from multiple sources and not every hit deserves the same level of trust. The new ‘Assign Confidence Levels’ scripting node for Thermo Scientific Compound Discoverer software was built to solve exactly that problem by bringing structured annotation confidence scoring directly into the workflow.

Abstract blue molecular structure with interconnected shapes and glowing particles.

The node is introduced in a recent publication¹ to bring a Schymanski-based annotation confidence frameworkinside Compound Discoverer software. A Schymanski-based annotation confidence framework gives scientists a standardized way to communicate how strongly the available MS evidence supports a proposed compound annotation².

Five level hierarchy of compound identification confidence in mass spectrometry adapted from Schymanski et al., 2014.

Confidence levels for compound identification in mass spectrometry, displayed from highest to lowest confidence (Level 1 at the top to Level 5 at the bottom).

Level 5 represents exact mass measurements obtained from MS and generated using the confidence scoring scripting node. Level 4 assigns a molecular formula based on MS isotope patterns and adduct information, supported by predicted compositions. Level 3 provides putative candidate identities derived from MS and MS² experimental data and database resources such as LipidSearch, ChemSpider, BioCyc, and other mass lists. Level 2 indicates a probable structure determined through MS and MS² spectral matching against library resources such as mzCloud. Level 1 represents confirmed structure identification, achieved using MS, MS², retention time, and comparison to an authentic reference standard, providing the highest level of confidence. Figure adapted from Schymanski et al. (2014)¹.

The developers describe a standalone post-processing node that extends the original five confidence levels to better separate cases that would otherwise be grouped too broadly. It uses evidence from existing identification nodes including Predict Compositions, Search mzCloud, Search mzVault, and Search ChemSpider.

What makes this especially useful is that it fits naturally into how experienced Compound Discoverer software users already work. This implementation packages custom post-processing actions on result tables into a purpose-built node that looks and behaves like part of the existing workflow.

Compound table from Compound Discoverer software showcasing the Assign Confidence Levels node output

The output is immediately practical. The node adds ConfLevel, ConfLevel_Flags, and Redundant Annotation Flag columns to the Compounds table, and it also creates a ConfidenceLevelSummary table that tallies compounds by confidence level. The setup guide also shows that key scoring settings, including ppm and match-factor thresholds, are exposed in the node parameters, which means users can review and tune the logic instead of treating the score as a black box.

For research scientists who already understand untargeted workflows, that is the real value. This node does not replace careful interpretation, but it does make interpretation easier to communicate. By converting mixed evidence into a consistent confidence framework, it improves reporting transparency and consistency across studies, and it gives teams a more standardized way to compare results between projects, analysts, and laboratories.

In practice, that should make several common tasks easier. When you are triaging candidates for standards follow-up, deciding which compounds are strong enough to highlight in a figure, or reviewing a result set with collaborators, a structured confidence field is far easier to defend than a loose combination of library hits and notes. It is also helpful that the node can be used during reprocessing, so existing result files can gain these additional confidence columns without rerunning the full analysis from scratch. Getting the node is straightforward. The mycompounddiscoverer.com annotation confidence scoring page currently offers version 1.52 downloads for Compound Discoverer software versions 3.3, 3.4, and 3.5, along with the installation guide. That makes adoption simple for labs already running one of the current 3.x releases.

Full installation instructions are on mycompounddiscoverer.com.

The best part of this release is that it addresses a very real gap in untargeted analysis. Scientists have become increasingly good at generating annotations, but reporting confidence still often depends on manual interpretation and inconsistent language. This node brings that step inside Compound Discoverer, where it becomes visible, reviewable, and reproducible. For teams already using Compound Discoverer software in small molecule workflows, that makes it a practical upgrade worth trying. Download it from mycompou nddiscoverer.com add it to your next workflow.