Quantitative structure retention relationship (QSRR) modelling for Analytes' retention prediction in LC-HRMS by applying different Machine Learning algorithms and evaluating their performance.

In metabolomics, retention prediction methods have been developed based on the structural and physicochemical characteristics of analytes. Such methods employ regression models, harnessing machine learning algorithms mapping experimentally derived retention time (tR) analytes with various structural and physicochemical descriptors, known as Quantitative Structure Retention Relationships (QSRR) models. In the present study, QSRR models have been developed by applying four Machine Learning regression algorithms, i.e. Bayesian Ridge Regression (BRidgeR), Extreme Gradient Boosting Regression (XGBR) and Support Vector Regression (SVR) using both linear and non-linear kernels, all tested and compared for their retention prediction ability on experimentally derived and on publicly available chromatographic data, using Molecular Descriptors to describe the physical, chemical or structural properties of molecules. Various configurations of the available datasets, in terms of the highly-correlated features levels (defined as the maximum absolute value of the Pearson's correlation coefficient calculated between any pair of features) they contained, were analyzed in parallel. This is the first study, to the best of our knowledge, of the effect of collinearity on the performance of QSRR predictive models. In the vast majority of cases studied there was no statistically significant difference in the performance of the generated QSRR predictive models among the specified dataset configurations, indicative of the ability of the selected regression algorithms to effectively handle collinearity. In terms of the individual performance of the selected regression algorithms, no pattern was found where one algorithm (or class of algorithms) stood out significantly relative to the others among the study datasets.

Platelet-activating factor acetylhydrolase and transacetylase activities in human plasma low-density lipoprotein.

In this study, we demonstrate the presence of a transacetylase activity in human plasma low-density lipoprotein (LDL) that transfers short-chain fatty acids from platelet-activating factor (PAF) and its close ether- and ester-linked analogues to ether/ester-linked lysophospholipids (lyso-PL). We show evidence that both PAF acetylhydrolase (PAF-AH) and transacetylase activities are inhibited to the same extent by serine esterase inhibitors, are resistant to heat treatment, and exhibit identical distributions in lipoprotein classes and in LDL subfractions. Additionally, the competitive inhibition of PAF-AH by lyso-PL, and the evidence that the recombinant PAF-AH also showed a similar transacetylase activity, suggest that PAF-AH is responsible for both activities. Using PAF as a donor molecule and lyso-PAF (1-O-alkyl-sn-glycero-3-phosphocholine) as an acceptor, the transacetylase activity showed typical allosteric kinetics, due to the positive co-operativity of the substrates, with apparent Vmax=19.6+/-3.4 nmol/min per mg of protein, apparent h=2.0+/-0.3 and apparent [S]0.5=9.4+/-2.3 microM at saturation for the concentration of lyso-PAF. The substrate specificity of the donor molecules was decreased by increasing the chain length of the acyl moiety in the sn-2 position of the glycerol. The ether linkage in the sn-1 position of the substrate was 30% more effective than the ester bond; cholesteryl acetate was inactive as an acetyl donor. The two acceptors tested, lyso-PAF and the ester-linked lyso-PC (1-acyl-sn-glycero-3-phosphocholine), showed similar specificity. Addition of exogenous lyso-PAF induced the transient formation of PAF-like aggregating activity predominantely in small dense LDL subfractions upon oxidation. We conclude that PAF-AH possesses both transacetylase and acetylhydrolase activities which remove PAF and its ether-linked analogues from LDL particles upon LDL oxidation. However, in atherogenic small dense LDL-5 particles, the transacetylase activity may acetylate extracellular lyso-PAF into biologically active PAF.

Involvement of phospholipids in apolipoprotein B modification during low density lipoprotein oxidation.

An increased amount of phospholipids remained attached on delipidated apolipoprotein B originated from oxidized low density lipoprotein (LDL). 31P nuclear magnetic resonance analysis of such apolipoprotein showed an organic phosphorus peak at -0.55 ppm, which suggests the formation of adducts (most probably Schiff bases) of oxidized phospholipids with apolipoprotein B. The above reaction occurs in parallel with the hydrolysis of oxidized phospholipids, catalyzed by the LDL-attached platelet-activating factor acetylhydrolase, and may contribute to the proatherogenic effect of oxidatively modified low density lipoprotein.

Copper-catalyzed oxidation mediates PAF formation in human LDL subspecies. Protective role of PAF:acetylhydrolase in dense LDL.

Free radical-mediated oxidation of cholesterol-rich LDL plays a key role in atherogenesis and involves the formation of oxidized phospholipids with proinflammatory biological activity. We evaluated the production of platelet-activating factor (PAF), a potent inflammatory mediator, in human LDL subspecies on copper-initiated oxidation (4 mumol/L CuCl2, 80 micrograms/mL for hours at 37 degrees C). PAF formation was determined by biological assay of HPLC-purified lipid extracts of copper-oxidized lipoproteins; chemical identity was confirmed by gas chromatographic and mass spectrometric analyses. PAF, characterized as the C16:0 molecular species, was preferentially produced in intermediate LDL (d = 1.029 to 1.039 g/mL) (8.6 +/- 5.7 pmol PAF/3 h per mg LDL protein) and light LDL (d = 1.019 to 1.029 g/mL), but was absent from dense LDL particles (d = 1.050 to 1.063 g/mL). As PAF:acetylhydrolase inactivates PAF and oxidized forms of phosphatidylcholine, we evaluated the relationship of lipoprotein-associated PAF:acetylhydrolase to PAF formation. We confirmed that PAF:acetylhydrolase activity was elevated in native, dense LDL (41.5 +/- 9.5 nmol/min per mg protein) but low in LDL subspecies of light and intermediate density (d 1.020 to 1.039 g/mL) (3.5 +/- 1.6 nmol/min per mg protein) [Tselepis et al, Arterioscler Thromb Vasc Biol. 1995;15:1764-1773]. On copper-mediated oxidation for 3 hours at 37 degrees C, dense LDL particles conserved 20 +/- 14% of their initial enzymatic activity; in contrast, PAF:acetylhydrolase activity was abolished in light and intermediate LDL subspecies. Clearly, the elevated PAF:acetylhydrolase activity of dense LDL efficiently diminishes the potential inflammatory role of endogenously formed PAF; nonetheless, formation of proatherogenic lysophospholipids results. In contrast, LDL particles of the light and intermediate subclasses can accumulate PAF on oxidative modification.

Distribution of PAF-acetylhydrolase activity in human plasma low-density lipoprotein subfractions.

The distribution of PAF-acetylhydrolase (PAF-AH) activity in 3 LDL subfractions prepared by density gradient ultracentrifugation as well as the rate of phosphatidylcholine (PC) hydrolysis during oxidation was studied. PAF-AH activity, measured before oxidation, was much higher in LDL3 subfraction (28.4 +/- 8.6 nmol/mg per min) comparing to LDL2 (14.1 +/- 5.8 nmol/mg per min), and to LDL1, 8.7 +/- 3.7 nmol/mg per min. During oxidation, the enzyme activity was continuously decreased and this phenomenon was more pronounced in LDL1. PC hydrolysis was studied measuring the lyso-PC production expressed as lyso-PC/Sph molar ratio. Before oxidation, the lyso-PC/Sph molar ratio, did not differ significantly among the LDL subfractions, whereas, 4 h after the onset of oxidation, it was significantly higher in LDL2 and LDL3 subfractions (0.42 +/- 0.12 and 0.45 +/- 0.10, respectively), comparing to LDL1 (0.29 +/- 0.06). Our results show that the distribution of PAF-AH activity in LDL subfractions is heterogeneous (mainly distributed in LDL2 and LDL3 subfractions) and it is positively correlated with higher lyso-PC production in those subfractions during oxidation. The contribution of this phenomenon to the enhanced susceptibility to oxidation as well as to the higher atherogenicity of the dense LDL subfractions is under investigation.

Platelet-activating factor formation during oxidative modification of low-density lipoprotein when PAF-acetylhydrolase has been inactivated.

A PAF aggregating activity corresponding to 427 +/- 91, 668 +/- 111 and 1319 +/- 217 pg/mg protein was detected when LDL was preincubated at pH 3.5 or with 4 mM PMSF or both for 30 min (treatments that inactivate PAF-AH) and then oxidized with 20 microM Cu2+ at 37 degrees C for 24 h. This molecule was characterized as PAF by its chromatographic behavior on TLC and other established methods and was further characterized as 1-O-hexadecyl-2-acetyl-sn-glycero-3-phosphocholine (C16: PAF) by its retention time on reverse phase HPLC and by fast atom bombardment-mass spectroscopy. Native LDL incubated under non oxidizing conditions, even when PAF-AH has been inactivated, or oxidized in the absence of PAF-AH inactivating agents or after pretreatment with 0.5 mM pBPB, does not produce detectable amounts of PAF. The kinetics of PAF formation in relation to PAF-AH activity, show that the apparent rate of PAF formation as well as its total amount depends on both the existence of oxidative conditions and the remaining PAF-AH activity the first hours following the onset of oxidation. Peroxidation of the phosphatidylcholine (PC) content of native LDL produces PAF-like aggregating activity much lower than that produced when intact LDL is oxidized and is not inhibited by BN 52021 as effectively as PAF produced by LDL peroxidation. Our results provide evidence that C16: PAF is formed during LDL peroxidation when PAF-AH has been inactivated and it does not result as a product of peroxidation of the LDL-PC content.