Aiello Donatella, Lucà Federica, Siciliano Carlo, Frati Paola, Fineschi Vittorio, Rongo Rocco, Napoli Anna
Department of Chemistry and Chemical Technologies, University of Calabria, Rende 87036, Italy.
Department of Biology, Ecology and Earth Science, University of Calabria, Rende 87036, Italy.
J Proteome Res. 2021 May 7;20(5):2607-2617. doi: 10.1021/acs.jproteome.0c01038. Epub 2021 Apr 27.
An analytical strategy for a matrix-assisted laser desorption mass spectrometry-based untargeted metabolomic study on vitreous humor (VH) was developed, looking for statistically significant parameters correlated to death time estimation. Five incubation stages of VH, 0, 24, 48, 72, and 96 h, at physiological pH and controlled temperature, were adopted to monitor time-dependent changes and correlate them with the postmortem interval (PMI). Using two multivariate statistical approaches, principal component regression (PCR) and partial least squares regression (PLSR), the PMI was assessed, considering the / values from mass spectra and the incubation time (ISt) as predictors. An independent validation set was used to evaluate the predictive capability of the models through the coefficient of determination () and the root-mean-square error (RMSE). Different pre-treatments were applied to the raw mass spectra, and their performance in assessing PMI was evaluated. Based on the best outcomes in terms of both and RMSE, multiplicative scatter correction combined with a logarithmic transformation was chosen. The results of PCR and PLSR based on the selected pre-treatment are encouraging because validation is about 0.95 for both models. Moreover, the prediction error is 6 h for both models, when PMI is lower than 1 day. Although these results are obtained by the uncritical application of the models, they are comparable to or even better than those reported in the literature. Notwithstanding, we consider that many influences, such as passive diffusion, functional loss of tissues, and advanced autolytic processes, could not get captured . However, the developed approach was optimized using VH samples and overcomes the limitations of the vast majority of methods that require validation for serum and/or urine samples.
开发了一种基于基质辅助激光解吸质谱的非靶向代谢组学研究玻璃体液（VH）的分析策略，以寻找与死亡时间估计相关的具有统计学意义的参数。采用在生理pH值和可控温度下VH的五个孵育阶段，即0、24、48、72和96小时，来监测随时间的变化并将其与死后间隔时间（PMI）相关联。使用主成分回归（PCR）和偏最小二乘回归（PLSR）这两种多元统计方法，将质谱中的/值和孵育时间（ISt）作为预测因子来评估PMI。使用独立验证集通过决定系数（）和均方根误差（RMSE）来评估模型的预测能力。对原始质谱应用了不同的预处理，并评估了它们在评估PMI方面的性能。基于在和RMSE方面的最佳结果，选择了乘性散射校正结合对数变换。基于所选预处理的PCR和PLSR结果令人鼓舞，因为两个模型的验证均约为0.95。此外，当PMI低于1天时，两个模型的预测误差均为6小时。尽管这些结果是通过对模型的不加批判的应用获得的，但它们与文献中报道的结果相当甚至更好。尽管如此，我们认为许多影响因素，如被动扩散、组织功能丧失和晚期自溶过程，可能无法被捕捉到。然而，所开发的方法使用VH样本进行了优化，克服了绝大多数需要对血清和/或尿液样本进行验证的方法的局限性。
