Introduction: Serum metabolomics, by representing the pathophysiological state of the organism, have been applicable to biomarker discovery in complex clinical status. Despite the great promises of metabolomics, unfortunately, conventional analytical platforms, such as the ones based on ultra-high-performance chromatography associated with high resolution mass spectroscopy (UHPLC-HRMS), present limitations when applied to large-scale populations due to their associated high costs, analytical complexity, and laborious and time-consuming nature. The current work compared an UHPLC-HRMS platform to an alternative platform based on FTIR spectroscopy to acquire the metabolic fingerprint of the system to predict mortality at an intensive care unit (ICU). Methods: Predictive models of mortality of 16 patients, from which half died at the ICU, were developed based on serum metabolomics acquired by UHPLC-HRMS and FTIR spectroscopy. Models were developed based on principal component analysis followed by linear discriminant analysis (PCA-LDA) from serum acquired 7 days before the clinical outcome, i.e., deceased or ICU discharge. Results: After feature selection, it was possible to develop excellent models based on a set of serum metabolites with 100% accuracy for the validation dataset. It was also possible to develop excellent models, based on the molecular fingerprint obtained by FTIR spectroscopy, that, after optimization of spectral pre-processing and band selection, resulted in a 92% accuracy for the validation dataset. Conclusions: The FTIR spectroscopy platform enabled the development of predictive models as good as the ones developed using data obtained from the UHPLC-HRMS platform, while being applicable in rapid, simple, more economical, and high-throughput analysis. All these characteristics are of paramount importance for application in large-scale populations, and consequently, the validity of the predictive models.

Acknowledgements: This work was supported by the DSAIPA/DS/0117/2020, PTDC from the Portuguese Foundation for Science and Technology, and by the IPL/IDI&CA2024/R-DICIP_ISEL financed by Instituto Politécnico de Lisboa.