How does the FoldGO method work?

The FoldGO algorithm sorts the input data by the fold change values, divides it into quantiles and generates various combinations of neighboring quantiles (fold change intervals). At the first step FoldGO performs functional enrichment analysis for the differentially expressed genes set. Then, at the second step, FoldGO calculates the overrepresentation of GO terms (preselected at the first step) in genes responding within each fold-change interval compared to the whole list of differentially expressed genes. GO terms that showed significant enrichment for one or more fold-change intervals under multiple testing correction are considered as fold-change-specific. The method was previously described and validated in Omelyanchuk et al (2017). FoldGO is also available as an R package via Bioconductor (Wiebe, 2018).

Which species does FoldGO support?

Currently, FoldGO allows processing data for the following species: Arabidopsis thaliana (TAIR ID), Drosophila melanogaster (Entrez ID), Homo sapiens (Entrez ID, Ensembl ID), Mus musculus (Entrez ID), Сaenorhabditis elegans (Entrez ID). The species list is to be updated. Note that FoldGO supports a limited set of gene identifiers (listed above). If your data contains gene identifiers which are not supported by FoldGO, you can convert them using third-party tools, e.g. DAVID Gene ID Conversion Tool (How to convert gene ID's using DAVID).

Which input data can I use?

As input data FoldGO requires two tab-separated or comma-separated tables for up- and down-regulated genes in which the first column contains Gene IDs and the second one contains the corresponding log transformed fold-change values. If fold-change values in your data are not log transformed, they can be easily converted to such in MS Excel (or similar) (log transformation in MS Excel) or in R via the log2 function (log transformation in R).

GeneID	logFC
AT3G65420	3.6
AT1G78450	1.5
AT2G66890	2.1
...	...

Additionally, the user can specify the background set of genes e.g. all genes which have corresponding probes on DNA microarray or gene universe from RNA-seq experiment. A background set of genes must be presented as a simple list of gene IDs:

AT1G01120

AT1G02520

AT1G02730

AT1G02850

AT1G02900

...

How to run the analysis?

To analyze your data, load two files from your computer, separately for up- and down-regulated genes, then press the “run” button.

One can test FoldGO using an example data by pressing the “Load example” button.

Advanced parameters: what can I customize?

• Number of quantiles - one can define the number of quantiles the input sets of genes should be separated on (the default is 6).
• FDR/FWER for step 1- the threshold to preselect GO terms from DEG set annotation for fold-change-specific enrichment. By default, it equals to 1 so FoldGO analyzes all GO terms annotated to an input set of genes. But you can make it more strict (e.g. 0.05) to filter the GO terms, that are overrepresented in the whole list of differentially expressed genes.
• FDR/FWER for step 2 - the threshold for selection of fold-change-specific GO terms (the default is 0.05).
• Correction method for step 1/2 - Multiple testing correction method for GO terms preselection and fold-change-specific enrichment analysis. FoldGO provides various methods for the correction on multiple testing, such as Benjamini-Hochberg, Benjamini-Yekutieli for FDR and Bonferroni for FWER. Although, FoldGO allows to choose different correction methods for different steps of analysis, we highly recommend to use the same method for both steps.
• Minimal number of genes annotated to a term - defines the minimal number of genes to be annotated to a GO term so that it is taken into the analysis (the default is 1). Can be used to filter well-described GO terms.

Results: GO profiles

Fold-change-specific GO Profile shows GO terms significantly associated with certain fold change intervals. Fold change intervals are depicted as yellow and blue boxes for up- and down-regulated genes, correspondingly. Numbers on the horizontal axis denote the boundary log fold change values for quantiles. The diagram presents only fold-change-specific GO terms. If the gene was associated fold-change-specifically with down-regulation but not fold-change-specifically with up-regulation (or vise versa) than the non-fold-change-specific interval (the default is 1-6) will be also shown.

Results: graphs

The Directed Acyclic Graph (DAG) is a most common way to visualize Gene Ontology data. Here graphs are tailored to represent fold-change-specific GO terms together with relations between GO terms. Note that if all ancestor GO terms of fold-change-specific GO term are not fold-change-specific they will not be shown. On the graph, fold-change-specific GO terms are shown as vertices of yellow color in case of up-regulation and blue in case of down-regulation. The graph is interactive so one can move vertices and look through additional information about the fold-change-specific GO terms by clicking on the corresponding vertices. Also you can zoom in/out the graph using the mouse wheel.

One can get more information about fold-change-specific GO term such as statistics and fold-change interval by just pointing the cursor on the corresponding vertex.

Results: tables

Tables contain the most complete information about GO terms under study. The FoldGO provides separate tables for fold-change-specific and not fold-change-specific GO terms and for up- and down-regulation. Columns contain the following information:

• Ids - a GO term ID according to Gene Ontology Consortium
• Namespace - a namespace to which the GO term belongs. Can be one out of three main ontologies: biological process, molecular function or cellular component.
• Name - a GO term full name.
• Interval (fold change) - a quantile or a quantile range on which the GO term showed the most significant enrichment (boundary fold change values for the quantile or quantile range).
• DEG GO enrichment stats - statistics for the first step of the analysis (GO terms preselection) which is equal to the conventional GO enrichment analysis, e.g. Singular Enrichment Analysis (SEA).
• Fold-change-specific enrichment stats - statistics from the fold-change-specific enrichment analysis (step 2). Here p- and q-values are provided for the interval which showed the most significant association with the corresponding GO term.

What is example data?

As an example dataset we use the data derived from RNA-seq experiment on phytohormone auxin treatment of Arabidopsis thaliana roots. The details on the experimental design and the conclusions about fold-change-specific analysis were described in (Omelyanchuk et al 2017).

Here are the links for the data used in the example:

up-regulated genes: link

down-regulated genes: link

background set of genes: link