LipidSig: a web-based tool for lipidomic data analysis
Correlation analysis
In this section, we provide a comprehensive correlation analysis to assist researchers to interrogate the clinical features that connect to lipids species and other mechanistically relevant lipid characteristics.
Correlation analysis between lipids and clinical features is broadly used in many fields of study, such as Bowler RP et al. discovering that sphingomyelins are strongly associated with emphysema and glycosphingolipids are associated with COPD exacerbations.
Hence, continuous clinical data can be uploaded here, and diverse correlation analyses are offered. For instance, the Correlation Coefficient and Linear Regression are supported for continuous clinical data. Moreover,
lipids can be classified either by lipid species or by lipid categories when conducting these correlation analyses.
Data Source
Demo dataset
Bowler, Russell P., et al. "Plasma sphingolipids associated with chronic obstructive pulmonary disease phenotypes."
American journal of respiratory and critical care medicine 191.3 (2015): 275-284.
This paper detected 69 distinct plasma sphingolipid species in 129 current and former smokers by targeted mass spectrometry.
This cohort was used to interrogate the associations of plasma sphingolipids with subphenotypes of COPD including airflow obstruction, emphysema, and frequent exacerbations.
For the data sources, users can either upload their datasets or use our Demo datasets.
The datatype of the dataset must be continuous data. The dataset needs to contain two tables,
‘lipid expression data’ and ‘condition table’. Another optional tables for adjustment and lipid characteristics analysis are also can be uploaded by users.
The Correlation Coefficient gives a summary view that tells researchers whether a relationship exists between clinical features and lipid species,
how strong that relationship is and whether the relationship is positive or negative. Here we provide three types of correlations, Pearson, Spearman, and Kendall,
and adjusted by Benjamini & Hochberg methods. The cut-offs for correlation coefficient and the p-value can be decided by users.
A heatmap will show after users inputting cut-offs and choosing a value for clustering/methods for clustering.
Users can use either correlation coefficient between clinical features (e.g. genes) and lipid species or choose their statistic instead.
Linear regression is a statistical technique that uses several explanatory variables to predict the outcome of a continuous response variable,
allowing researchers to estimate the associations between lipid levels and clinical features.
For multiple linear regression analysis, additional variables in ‘adjusted table’ will be added into the algorithm and used to adjust the confounding effect.
Once calculation completes, each lipid species will be assigned a beta coefficient and t statistic (p-value), which can be chosen for clustering.
The Correlation Coefficient gives a summary view that tells researchers whether a relationship exists between clinical features and user-defined lipid characteristics,
how strong that relationship is and whether the relationship is positive or negative. Here we provide three types of correlations, Pearson, Spearman, and Kendall,
and adjusted by Benjamini & Hochberg methods. The cut-offs for correlation coefficient and the p-value can be decided by users.
A heatmap will show after users inputting cut-offs and choosing a value for clustering/methods for clustering.
Users can use either correlation coefficient between clinical features (e.g. genes) and lipid characteristics or choose their statistic instead.
Multiple linear regression is a statistical technique that uses several explanatory variables to predict the outcome of a response variable,
allowing researchers to estimate the associations between lipid levels and clinical features (i.e., genetic polymorphisms).
In this page, the lipids will be classified by the user-selected lipid characteristics (e.g. class), then implementing multiple linear regression analysis.
Each variable (the pair of lipid characteristics and clinical features) will be assigned a beta coefficient and t statistic (p-value), which can be chosen for clustering.