Feature detection with controlled error rates in LC/MS images

Area of Science:

• Proteomics
• Analytical Chemistry
• Computational Biology

Background:

• Liquid Chromatography/Mass Spectrometry (LC/MS) is crucial for peptide identification.
• Existing Median M-N rule algorithms lack robust statistical error control.
• Accurate feature detection is vital for reliable LC/MS data analysis.

Purpose of the Study:

• To extend the Median M-N rule for statistical error control in peptide detection.
• To introduce a novel algorithm, the Quantile M-N rule, for improved feature detection.
• To provide statistical bounds for false-positive and false-negative rates in LC/MS data.

Main Methods:

• Developed an extension of the Median M-N rule to calculate a statistical bound for the false-positive rate.
• Investigated the false-negative rate associated with the M-N rule.
• Analyzed the types of signals detectable and the limit of detection.

Main Results:

• The Quantile M-N rule provides statistical control over false-positive rates.
• Insights into the false-negative rate and detection limits were established.
• The algorithm demonstrates applicability to various feature detection methods.

Conclusions:

• The Quantile M-N rule offers enhanced statistical rigor for peptide signal detection in LC/MS.
• This method improves the reliability of feature detection by controlling both false-positive and false-negative rates.
• The Quantile M-N rule is a versatile tool applicable to a broad range of feature detection algorithms.