MRH test dataset

 

Overview This dataset was generated to evaluate and benchmark the Multiresolution Hierarchical (MRH) peak detection algorithm. It consists of high-resolution mass spectrometry (HRMS) data from controlled mixtures of chemical standards, providing a reliable basis for assessing detection sensitivity, localization accuracy, and the effectiveness of confidence metrics like the Ambiguity Score (Q).

 

Sample Composition and Preparation The dataset is derived from the INTERFLAB experiment. It comprises in-house generated mixtures of flame retardants prepared at known proportions. To facilitate a robust sensitivity analysis, the samples were prepared across a wide range of predefined concentrations: 0.1, 0.5, 1.0, 5.0, 10.0, 25.0, 50.0, 100.0, 250.0, 500.0, and 1000.0 ng/ml.

 

Data Acquisition and Instrumentation Measurements were performed using a Thermo Fisher Scientific Q Exactive GC Orbitrap GC-MS/MS system. Key acquisition parameters include:

• Mode: Profile mode (essential for MRH’s direct raw-data processing).

• Acquisition Rate: High-acquisition rate.

• Mass Accuracy & Resolution: Data was acquired at high resolution.

1. Raw Data (.mzML): These files contain the raw, uncentroided profile data for each concentration level. They preserve the full measurement precision required for hierarchical decomposition.

2. Ground Truth Lists: These files provide the curated list of "true" peaks for each concentration.

    ◦ Curation Process: Predicted peak positions were matched against the raw data, and the closest local maxima (ΔRT=0.2 s, ΔMZ=0.1 Da) were designated as true peaks to ensure alignment between predicted and observed values. Figure of matched_regions are provided for a visual check.

    ◦ Data Availability Labels: Instances where no raw measurements could be associated with predicted peaks (typically at the lowest concentrations) are explicitly marked as "no data".

 

Potential Applications This dataset is intended for researchers developing feature detection algorithms for HRMS. It is particularly valuable for testing performance and for validating the localization precision of new peak-picking methodologies