75
4.4.3.2 Statistical Data Analysis
Lipidomic datasets usually comprise tens to hundreds of identified lipid species,
along with a large number of unidentified species, and thus the compound identification is followed by statistical analysis for validation of data, depending on the biological question of interest. The integrated collaboration of lipid profiling and uni-/
multivariate statistics in a lipidomic approach helps in discovering potential lipid
biomarkers and in-depth understanding of lipid biochemical pathways. The univariate method such as analysis of variance (ANOVA) is often used for the comparison
of two or more groups. One drawback of univariate approach is that it may result in
too many false-positive findings due to many hypothesis tests being performed (Han
et al. 2012; Hendriks et al. 2011; Vaz et al. 2015). Multivariate approaches such as
multivariate analysis of variance (MANOVA), permutational multivariate analysis
of variance (PERMANOVA), and similarity percentages (SIMPER) are often used
for validation of proposed models. The principal component analysis (PCA) and
partial-least-squares discriminant analysis (PLS-DA) consider the correlation structure of the data and reduce its dimensionality by constructing so-called latent variables, which are combinations of the original variables, and also facilitating the
visualization of the data in two or three dimensions. PLS-DA is mainly used for
biomarker discovery and PCA or cluster analysis for obtaining information about the
(separation of) groups of samples and/or metabolites (Chen et al. 2016; Hendriks
et al. 2011; Kumari et al. 2014b; Melo et al. 2015; Vaz et al. 2015).
4.4.3.3 Bioinformatic Interpretation and Pathway Analysis
The next step is development of bioinformatic and systems biology approaches to
link the changes of cellular lipidome to alterations in the biological functions,
including the enzymatic activities that are involved in the biosynthesis of the altered
lipid classes and molecular species, addressing the biological question of interest.
The lipid classes and individual molecular species involved in the biosynthesis of a
particular lipid class are clustered, and the known biosynthesis and/or remodeling
pathways are utilized to simulate the ion profiles of the lipid class of interest (Haimi
et al. 2006; Wenk 2010; Wang et al. 2016 and references therein). A best match
between the simulated and determined ion spectra is achieved from simulation
based on the known pathways. Numerous parameters involved in the biosynthetic
pathways can be derived from the simulation of high-throughput lipidomic datasets
that can be used for quantitative interpretation of the pathways involved in adaptive
changes in lipid metabolism after any perturbation (Han et al. 2012; Vaz et al. 2015;
Wenk 2010; Wang et al. 2016). To facilitate the biological interpretation of lipidomic data, the identified metabolites are integrated and visualized in the context of
metabolic pathways obtained from public databases (Haimi et al. 2006; Han et al.
2012; Sreenivasaiah et al. 2012; Vaz et al. 2015; Wenk 2010). Further, a statistical
approach toward pathway analysis involving metabolite set enrichment analysis
4 Seaweed Lipidomics in the Era of ‘Omics’ Biology: A Contemporary Perspective
4.4.3.2 Statistical Data Analysis
Lipidomic datasets usually comprise tens to hundreds of identified lipid species,
along with a large number of unidentified species, and thus the compound identification is followed by statistical analysis for validation of data, depending on the biological question of interest. The integrated collaboration of lipid profiling and uni-/
multivariate statistics in a lipidomic approach helps in discovering potential lipid
biomarkers and in-depth understanding of lipid biochemical pathways. The univariate method such as analysis of variance (ANOVA) is often used for the comparison
of two or more groups. One drawback of univariate approach is that it may result in
too many false-positive findings due to many hypothesis tests being performed (Han
et al. 2012; Hendriks et al. 2011; Vaz et al. 2015). Multivariate approaches such as
multivariate analysis of variance (MANOVA), permutational multivariate analysis
of variance (PERMANOVA), and similarity percentages (SIMPER) are often used
for validation of proposed models. The principal component analysis (PCA) and
partial-least-squares discriminant analysis (PLS-DA) consider the correlation structure of the data and reduce its dimensionality by constructing so-called latent variables, which are combinations of the original variables, and also facilitating the
visualization of the data in two or three dimensions. PLS-DA is mainly used for
biomarker discovery and PCA or cluster analysis for obtaining information about the
(separation of) groups of samples and/or metabolites (Chen et al. 2016; Hendriks
et al. 2011; Kumari et al. 2014b; Melo et al. 2015; Vaz et al. 2015).
4.4.3.3 Bioinformatic Interpretation and Pathway Analysis
The next step is development of bioinformatic and systems biology approaches to
link the changes of cellular lipidome to alterations in the biological functions,
including the enzymatic activities that are involved in the biosynthesis of the altered
lipid classes and molecular species, addressing the biological question of interest.
The lipid classes and individual molecular species involved in the biosynthesis of a
particular lipid class are clustered, and the known biosynthesis and/or remodeling
pathways are utilized to simulate the ion profiles of the lipid class of interest (Haimi
et al. 2006; Wenk 2010; Wang et al. 2016 and references therein). A best match
between the simulated and determined ion spectra is achieved from simulation
based on the known pathways. Numerous parameters involved in the biosynthetic
pathways can be derived from the simulation of high-throughput lipidomic datasets
that can be used for quantitative interpretation of the pathways involved in adaptive
changes in lipid metabolism after any perturbation (Han et al. 2012; Vaz et al. 2015;
Wenk 2010; Wang et al. 2016). To facilitate the biological interpretation of lipidomic data, the identified metabolites are integrated and visualized in the context of
metabolic pathways obtained from public databases (Haimi et al. 2006; Han et al.
2012; Sreenivasaiah et al. 2012; Vaz et al. 2015; Wenk 2010). Further, a statistical
approach toward pathway analysis involving metabolite set enrichment analysis
4 Seaweed Lipidomics in the Era of ‘Omics’ Biology: A Contemporary Perspective