PERSONAL DE APOYO
BURDISSO Paula
congresos y reuniones científicas
Título:
Data fusion approach applied in chemometrics-assisted metabolomics analysis
Autor/es:
MARTINEZ BILESIO, ANDRÉS R.; PUIG CASTELLVI, FRANCESC; TAULER, ROMA; RASIA, RODOLFO M.; BURDISSO PAULA; GARCÍA REIRIZ, ALEJANDRO G.
Lugar:
Roma
Reunión:
Congreso; CAC2022: 18th Chemometrics in Analytical Chemistry Conference; 2022
Resumen:
Biofluid metabolomics offers a non-invasive approach which provides key insights into the status and dynamics of health and disease conditions. As in other omics sciences, metabolomics data are complex and require appropriate data analysis tools to extract the biologically relevant information which often is hidden in them. Most metabolomic studies use a single type of biofluid sample such as serum, urine, or feces, among others. However, the interpretation of the complex biological mechanisms can be boosted by simultaneously analysing different types of biological matrices from the same individual applying data fusion (DF) procedures. Multivariate Curve Resolution with Alternating Least Squares (MCR-ALS) is a method that aims to resolve and analyse mixed data. One strength of MCR-ALS is that it can be applied in different stages of data workflow for many purposes, including resonance integration, data compression and exploratory data analysis. In this work, a new guideline is proposed for performing the joint analysis of different metabolomic datasets (serum, urine, metadata). This guideline includes two strategies, DF1 and DF2, which use MCR-ALS in a consecutive way, coupling the results of one MCR-ALS analysis to the next. In the present work, DF1 and DF2 were applied on 1H RMN spectral data from urine and serum samples combined with biochemical metadata provided by 145 healthy volunteers. Both DF methodologies were able to successfully merge and process the multiblock datasets, allowing to visualize the data structure and to highlight the most relevant variables associated with the previously reported discriminating factors in healthy cohorts. Eventhough both methodologies made it possible to find differences in sex, only the DF2 strategy showed age-related components.