search.noResults

search.searching

dataCollection.invalidEmail
note.createNoteMessage

search.noResults

search.searching

orderForm.title

orderForm.productCode
orderForm.description
orderForm.quantity
orderForm.itemPrice
orderForm.price
orderForm.totalPrice
orderForm.deliveryDetails.billingAddress
orderForm.deliveryDetails.deliveryAddress
orderForm.noItems
18 August / September 2019


Gradient Retention Time Prediction for 653 Pesticides on a Biphenyl LC Column Using Machine Learning


by Leon P. Barrona* and Neil J. Loftusb


Shimadzu Corporation, Manchester, United Kingdom *Email: leon.barron@kcl.ac.uk; Tel. +44 (0)20 7848 3842


aDept. Analytical, Environmental & Forensic Sciences, King’s College London, Franklin Wilkins Building, 150 Stamford St., London SE1 9NH, United Kingdom b


The development of a machine learning model for the accurate prediction of 653 pesticide retention times (tR ) on a biphenyl stationary phase is


presented. Using an ensemble of four multi-layer perceptron neural networks, prediction of 75% of all compounds lay within 39 s of measured tR


=0.8555 for n=98 compounds).


over a 12 min gradient elution method. A total of 16 input variables were selected and included constitutional indices (number of double/ triple bonds, carbons and oxygen atoms) ring descriptors (number of 4-9 membered and benzene rings) and molecular properties (unsaturation index, hydrophilic factor, unsaturation index, logP and logD). Correlation of blind test data was excellent (R2


LogD contributed significantly more to predictions compared to other descriptors, but 6-membered/benzene ring and hydrophilicity descriptors contributed more on biphenyl than for separations on C18


and a wide applicability domain. The ability to accurately predict tR screening applications using an alternative selectivity to C18


Introduction


Qualitative and quantitative analysis of environmental samples containing large numbers of analytes in single LC-MS/MS assays has become more widespread in recent years. This has been rapidly enabled by the development of fast scanning and trapping-type mass spectrometry instruments and most notably with high resolution accurate mass spectrometry (HRMS). Despite having accurate m/z measurements in both full-scan and tandem MS modes, isomers often exist that make identification challenging for some compounds, especially in complex matrices. Chromatographic retention time (tR


) is usually used to further distinguish compounds, where standards are available. Unfortunately, this is not always the case. For pharmaceuticals and illicit drugs, for example, the presence of Phase I and II metabolites still pose a challenge for confirmation in silico as reference materials of high purity either are not available or are prohibitively expensive to procure. Also, retention on C18


media is limited for many such polar compounds.


Where multiple unknowns exist in a sample, the prediction of tR


may rapidly enable shortlisting of candidates. Retention


media. Principal component analysis of descriptor data showed good clustering overall on biphenyl media represents an excellent opportunity for in silico suspect , especially when coupled to high resolution mass spectrometry.


Figure 1. Multiple reaction monitoring chromatograms of 652 pesticides on the 100 x 2.1 mm, 2.7 µm Raptor biphenyl column. Note: dimethirimol data removed for clarity (measured tR


= 5.00 min; predicted tR


prediction has been the focus of significant research and has been particularly successful


in gas chromatography [1-4]. However, tR prediction in liquid chromatography (LC) has been more challenging. Mechanistic approaches, e.g., using linear solvation energy relationships, have been able to successfully predict retention of compounds using sets of measured tR


or retention factor


(k) data gathered under an array of different experimental conditions, such as mobile


phase composition, pH, temperature, flow rate, etc. However, the number of experiments generally required to build such models is often high and application to large numbers of (unknown) compounds, especially under gradient conditions has, on the whole, been very limited. Among other computational approaches, machine


learning has been used for many years for tR prediction of peptides [5]. Recently, machine learning has been used successfully for


= 4.88 min).


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58  |  Page 59  |  Page 60  |  Page 61  |  Page 62  |  Page 63  |  Page 64  |  Page 65  |  Page 66  |  Page 67  |  Page 68