Abstract
The use of retention time is often critical for the identification of compounds in metabolomic and lipidomic studies. Standards are frequently unavailable for the retention time measurement of many metabolites, thus the ability to predict retention time for these compounds is highly valuable. A number of studies have applied machine learning to predict retention times, but applying a published machine learning model to different lab conditions is difficult. This is due to variation between chromatographic equipment, methods, and columns used for analysis. Recreating a machine learning model is likewise difficult without a dedicated bioinformatician. Herein we present QSRR Automator, a software package to automate retention time prediction model creation and demonstrate its utility by testing data from multiple chromatography columns from previous publications and in-house work. Analysis of these data sets shows similar accuracy to published models, demonstrating the software’s utility in metabolomic and lipidomic studies.
Highlights
Mass spectrometry (MS) is commonly used for metabolite and lipid profiling
Like any statistical extrapolating the retention times observed the set to may lead to inaccurate results
Though the training set contained a wide of settraining may lead inaccurate results
Summary
Mass spectrometry (MS) is commonly used for metabolite and lipid profiling. MS allows the measurement of mass to charge rations (m/z) of hundreds of compounds in a single analysis. While incredibly useful, determining the identity of a compound only by its m/z can be difficult. The same m/z can belong to isobaric and isomeric compounds in the same organism [1,2,3,4] or can be artifacts caused by the MS [5]. While effective in a number of cases many metabolites have similar fragments. Metabolite MS/MS libraries and fragment prediction software are not yet sufficient to identify all compounds solely by fragmentation [1,2,6,7]. To ensure the proper identification of compounds, observed orthogonal measurements are needed.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call