TFactS Manual

CONTENTS:

DESCRIPTION
AUTHORS
INPUT

TFactS MODES
Supported organisms
Gene specification

ANNOTATED SIGNATURES

Data sources
Target gene signatures

METHOD

Comparison Statistics
Statistical significance
Example

PUBLICATIONS
ACKNOWLEDGEMENTS

DESCRIPTION

is designed to predict which transcription factors are regulated, inhibited or activated in a biological system based on lists of upregulated and downregulated genes generated in microarray experiments.
TFactS takes as input lists of up- and/or down-regulated genes (query genes), compares it with a catalogue of annotated target genes, and returns three lists of transcription factors whose annotated target genes show a significant overlap with the query genes. The first list shows the Regulated Transcription Factors(TF) using the the Sign-Less catalogue, the second list shows the activated TF and the third list shows the Respressed TF. Both the activated and repressed lists are produced using the Sign-Sensitive catalogue.

AUTHORS

Ahmed Essaghir & Jean-Baptiste Demoulin

INPUT

TFactS MODES:

TFactS can be used in 2 modes (Batch Mode and Simple Mode):

BatchTFactS is made to simplify multiple submissions:
- The user has to submit a zipped folder (only .zip is accepted at the moment).
- in the zipped folder there should be files (all the file names should alphanumeric with only .txt extension)
- each file.txt in the zipped folder should contain 2 tab-separated columns. The first column will contain the Up-regulated gene symbols (or Entrez gene Ids);
  the second column will contain the Down-regulated gene symbols (or Entrez gene Ids)
- For optimal usage: Please be sure that your HGNC gene sybmols (or Entrez gene Ids) are standard and do not contain characters like );/\!
- As an example of file.txt (the headers are not mandatory):
  
  UP DOWN
  
  Foxo1 Srebf1
  
  Foxo3 Ccna1
  
  Tp53 Sp1
  
  Cdkn1b Nfkb1
  
  Cdkn1a
TFactS (Simple Mode) takes as input one or two lists of query genes:
- UP: list of up-regulated genes
- DOWN: list of down-regulated genes
The user can specify which catalogue to use. "Add Data" button can be used to sepcify costum catalogues.
- user can upload his own catalogue
- user can add his catalogue to the TFatcS catalogue
- optionally user can make a suggestion of a new target gene list to be added to TFactS catalogue (after reviewing)
The user can specify how many random selection to use for negative control (RC : see Methods section)
The user can specify desired thresholds for positive control and False discovery control (see Methods section)
The user can limit the comparison list of TFs to the ones having more than a certain number of target genes (target gene # threshold)

UP	DOWN
Foxo1	Srebf1
Foxo3	Ccna1
Tp53	Sp1
Cdkn1b	Nfkb1
Cdkn1a

Supported organisms:

TFactS primarily aims at interpreting human data. Rat and mouse human orthologue genes are also supported.

Gene specification:

genes should be specified by their HGNC official names or by their Entrez ids.

ANNOTATED SIGNATURES

TFactS relies on catalogues of transcription factor signatures. A signature is defined here as the list of genes regulated by a transcription factor.
The Sign-Sensitive catalogue enumerates pairwise relationships between transcription factors (TF) and their target genes (TG), documented with the regulation type (up or down). The Sign-Less catalogue doesnt contain regulation type information.

Data sources

Lists of TG are obtained from literature (see references) and databases like , , curated and . We annotated some of these lists with the regulation type (up or down) based on the original publications (from PubMed). SREBP and p53 TG were based on our previous reports[1,2] and were complemented by data in TRED with the regulation type from original publications. The FOXOs and STATs transcription factors TG lists were built from published papers[3-21].

Target gene signatures

In our catalogue of TF target genes , we define a signature as a set of all the target genes of this TF. And a Generic TF is defined as a TF representing a very close family of transcription factors: STAT, e.g, is a Generic TF for STAT1, STAT3 and STAT5.

Specific signatures

A Specific Signature is the normal signature as defined above for each TF.

Generic signatures

A Generic Signature is the union of all Specific signatures of the TF represented as a Generic transcription factor.

METHOD

The program compares the list(s) of query genes (up and/or down) with a catalogue of signatures.

Comparison Statistics

For each TF Three hypotheses are tested using a contingency table as follows :

TF	User: Present	User: Absent	Total
Catalogue: Present	k	m-k	m
Catalogue: Absent	n-k	N+k-n-m	N-m
Total	n	N-n	N

m: # TG of TF
n: # submitted genes
N: # signatures in the catalogue
k differs according to the hypothesis to be tested as follows :

Regulation of TF (using Sign-Less catalogue):
k: # submitted genes and regulated by TF
Activation of TF (using Sign-Sensitive catalogue):
k: # submitted genes and regulated by TF matching the regulation type
Repression of TF (using Sign-Sensitive catalogue):
k: # submitted genes and regulated by TF matching the opposite regulation type

Statistical significance

TFacts gives as output three tables, each table gives statistics for each of the three hypotheses (Regulation (sign-less catalogue), Activation and Repression (sign-sentisitve catalogue) of a TF)

P-value : is computed using the Fisher's test with the Hypergeometric formula :

E-value : is the number of tests performed (T) times the p-value :

E-value represents the Expected number of false positive in multiple testing.

Q-value : is computed using R package "Qvalue" from Storey et al.[22-24], Storey defines the q-value as the minimum False Discovery Rate (FDR) that can be attained when calling a feature (here TF) significant. They estimated the q-value for a given p-value by the formula bellow :

where t is the threshold for P-values below which the TF is called significant.
According to Storey et al. : a q-value threshold can be phrased in paractical terms as the proportion of the significant features (here TF) that trun out to be false leads.

Benjamini and Hochberg FDR controlling Procedure: this procedure is implemented as described in Benjamini and Hochberg [25]:
1. Rank ascending the P-values of the tests;
2. Compute for each P-value the corresponding FDR control:
3. find the largest rank i for which the P-value <= FDR Control
4. all the tests with ranked P-values from 1 to i are considered as significant under this FDRt
Random Control (RC): is the percentage of which a TF is called significant under a certain e-value threshold after a random simulation of user lists in specified number of repetitions :

where λ represents the e-value threshold, and rep stands for repetitions.
RC represents a non parametric control of false positives.

Example

Total Submitted genes (UP + DOWN): 43;
FOXO3 has 66 TG in the sign-sensitive catalogue;
among the submitted genes, 11 matched FOXO3 TG;
among these 11 genes : 8 matched the regulation type and 3 mismatched the regulation type.
Total number of Tests : 24;
Number of Repetitions (random simulation): 100;

**Activation of FOXO3**
FOXO3	User: Present	User: Absent	Total
Catalogue: Present	8	58	66
Catalogue: Absent	35	2328	2363
Total	43	2386	2429

P-value = 1.40e-4;
E-value = 3.36e-3;
Q-value = 3.36e-3;
FDR control = 2.08e-3 (≥ P-value);
RC ≤1%

**Repression of FOXO3**
FOXO3	User: Present	User: Absent	Total
Catalogue: Present	3	63	66
Catalogue: Absent	40	2323	2363
Total	43	2386	2429

P-value = 2.15e-1;
E-value = 5.15e+0;
Q-value = 3.06e-1;
FDR control = 1.67e-2 (< P-value);
RC ≤1%

If we set thresholds like : P-value ≤ 0.05; E-value ≤ 0.05; Q-value ≤ 0.05; FDR ≤ 0.05 and RC ≤ 5% Then we can say that FOXO3 is activated in this system.
The regulation hypothesis is based on the sign-less catalogue and the contingency table is built as the same as above without taking into account the sign of the regulations.

PUBLICATIONS

Essaghir A, Toffalini F, Knoops L, Kallin A, van Helden J, Demoulin JB: Transcription factor regulation can be accurately predicted from the presence of target gene signatures in microarray gene expression data. Nucleic Acids Res. 2010 Jun 1;38(11):e120. Epub 2010 Mar 9. Essaghir A, Demoulin JB: A Minimal Connected Network of Transcription Factors Regulated in Human Tumors and Its Application to the Quest for Universal Cancer Biomarkers. Plos One 7 (6), 2012, e39666.

ACKNOWLEDGEMENTS

Huu-Chi Hoang has graciously helped to develop BatchTFactS.
Mariama Elbaroudi has graciuously helped in the last database update.