1. Introduction to NFFinder

NFFinder is an bioinformatics approach that combines experimental gene expression profiles and data from publicly available databases to identify conditions or experiments that induce similar or opposite gene expression patterns.

2. Workflow

Figure 1. Workflow

NFFinder starts with transcriptomic profiles obtained from patients or celullar cultures as INPUT. After screening the public databases (GEO, Cmap and DrugMatrix) NFFinder identifies particular conditions (diseases, drugs, treated cell lines, among others) and retrieves experiments with similar or opposite gene expression profiles that may suggest new hypothesis for drug repositioning by recovering connections between genes, drugs and diseases related to the same biological process.

The OUTPUT shows the comparisons between samples from retrieved experiments and additional information from authors contributing to publications referenced in the results. These experts focus their research on diseases sharing common expression profiles with the INPUT introduced.

3. How it works

3.1. Input

The user queries NFFinder through two lists of genes as INPUT, one list represents over-expressed genes and the other list under-expressed genes. NFFinder compares these two gene lists against the gene signatures previously obtained by NFFinder identifying gene signatures with similar or opposite expression patterns.

Selecting the Load example button, the up and down regulated genes obtained from differential expression analysis of benign dermal neurofibroma compared to normal Schwann cells from GSE 14038 data (Miller et al., 2009) are loaded.

After gene profile's introduction, different fields must be filled out to define the query:

Figure 2. Input page

3.2. Output

3.2.1. Results

The picture below shows the main results page. By clicking each one of these buttons the user can access the results and the different views summarizing those results. Moreover, the user can filter the results to visualize the profiles obtained from the database of his or her choice.1

Figure 3. Main page of results

3.2.2. Ranked table of NFFinder results retrieved from databases

Figure 4. Ranked table of the experiments obtained in the analysis.

The above table shows comparisons among samples from databases' experiments sorted by their relationship's strength with the query (score). The top of the list offers comparisons with a higher relationship with the query (direct or inverse) and information about each experiment, including the link to the page where data is stored.

Filters shown in the Options area help the user to select only those results in which she/he is interested.

By selecting the different experiments and then clicking “Filter to marked rows” only those experiments marked will remain on the pages. You can unmark the selection by clicking Unmark Marked Rows or visualize again all the experiments by clicking Reset all filters. Moreover, the right check-boxes allows you to filter by the condition in which the experiments were carried out.

The marked rows selected in the NFFinder results will still selected in pages Summary, DRUG experiments and DISEASE experiments, and inversely, selections in any of these three pages will appear selected in the NFFinder results page.

All data tables can be exported by clicking over them with the right mouse button.

Disclaimer: The relationship used by NFFinder to connect experiments with drugs and diseases is obtained using MetaMap on the information provided by the researchers in the databases.

MetaMap performs a semantic analysis of the description text field from each experiment and identifies concepts drawn from a controlled vocabulary. These concepts are assigned to each experiment as tags for drugs and diseases.

It is the user's job to check the actual relationship between each experiment and a particular drug or disease, especially when GEO is chosen as queried database.

3.2.3. Summary

The page summarize the results obtained in the analysis. The upper part of the above picture shows the number of comparisons related to drugs (left) or diseases (right) with high, medium or low score. The lower part of the picture shows the average score of the experiments obtained with the same tag of drug (left) or disease (right) plus the number of times it appears in the results.

Figure 5. Summary.

3.2.4. DRUGS related to experiments

The page presents the drugs associated to gene expression experiments. The length of the horizontal bars indicates the total number of gene expression experiments retrieved from databases and the bar color indicates the importance of the experiment within the results according to the score (low, medium and high).

Figure 6. Drugs associated to gene expression experiments.

Results can be queried filtering the score through the bar shown on the bottom. With the left lateral bar you get a closer view of the desired area or also by using the Zoom bottom after selecting particular bars to visualize. You can undo the closer visualization by clicking Undo zoom or unmark the bars by pressing Unmark the selected area.

The figure below shows the information obtained by clicking at any single bar. On the right, there are the names of the experiments related to the drugs. At the same time, there is a drugs table which details the samples compared in each experiment including a link to the appropriate experiment of each database.

Figure 7. An example of the functionality of the page.

3.2.5. DISEASES related to experiments.

The page shows the diseases associated to gene expression experiments. It has the same organization and functionality as the drugs related to experiments page (the previous section) but showing those experiments which are related to diseases.

Figure 8. Diseases associated to gene expression experiments.

The additional button shown in the bottom of the above picture Gene expression heatmaps button, takes the user to a second display showing the expression profiles of input up- and down- regulated genes in the experiments related to the selected disease (see more in Gene expression heatmap section).

3.2.6. Gene Expression HEATMAPS.

Gene expression heatmaps allow the user to check the expression of his input up- and down-regulated genes retrieved from GEO database experiments associated to diseases.

Figure 9. Gene expression heatmap.

3.2.7. DISEASE-DRUG Interactions.

Disease-Drugs interactions displays upper and lower treemaps for diseases and drugs, respectively.

Particular diseases or drugs can be selected in the left filter shown in the picture bellow or by directly clicking the name in the treemap. Drugs or diseases related with selected diseases or drugs, respectively, will be highlighted in the treemap. The buttons shown in the page bottom help the user to reset filters (Reset filters) and deselect boxes (Unmark selected area). Details about the experiments containing tags related to drugs and diseases may be queried in the table below the treemaps.

Figure 10. Disease-Drug Interactions.

3.2.8. EXPERTS related to experiments.

Experts displays a list of authors contributing to publications referenced in drug or disease experiments. The treemap ranks the experts'list according to the number of their research articles. Results can be filtered by score through the bar shown in the left.

The References table on the right contains the title of authors' research article including a link to Pubmed to get more in-depth information about experts' work.

Figure 11. Experts related to experiments.

1Only databases selected in the input page appear in the main results page.

4. Example: dNFSC vs. NHSC

In the following example we're trying to find possible repositioning candidates by locating in NFFinder's databases either treatable conditions that present a phenotype similar to dNFSC or drugs and compounds that might be capable of reverting it.

We start off with microarray data from a GEO series: GSE14038 ( Miller et al. 2009, Rahrmann et al. 2013) and we use LIMMA (Bioconductor) to obtain a set a of differentially expressed genes comparing dNFSC samples with NHSC (reference) samples. We use a list of up-regulated and a list of down-regulated genes as an input for NFFinder and perform two different searches.

Up-regulated genes:

  • 204359_AT
  • 212758_S_AT
  • 212764_AT
  • 225342_AT
  • 226834_AT
  • 227290_AT
  • 231227_AT
  • ADM
  • ALPK2
  • AMIGO2
  • AQP1
  • ARHGAP22
  • ATP10A
  • ATP9A
  • BASP1
  • BCAT1
  • C10ORF90
  • CACHD1
  • CBLB
  • CD55
  • CDH11
  • CDH13
  • CHST15
  • CHSY3
  • COL6A1
  • COL6A2
  • COL6A3
  • CPE
  • DACT1
  • DNM3OS
  • DOCK10
  • EBF1
  • EN1
  • FAM110B
  • FAM155A
  • FAP
  • FAS
  • FGFR1
  • FOXD1
  • GFPT2
  • GLIPR2
  • GRIA3
  • HAS2
  • HMGA2
  • HRH1
  • HS3ST3A1
  • HS3ST3B1
  • HTRA1
  • KAZN
  • KIAA1462
  • KIAA1549
  • KIAA1644
  • KRTAP1-5
  • LDB2
  • LOXL1
  • LPPR4
  • LPXN
  • LUM
  • LY96
  • LYN
  • MAST4
  • MEIS2
  • MGAM
  • MME
  • MYO1B
  • NPAS2
  • PAPSS2
  • PCDH18
  • PID1
  • PLAU
  • POU2F2
  • PRDM1
  • PRRX1
  • RAB3B
  • S100A4
  • SETBP1
  • SGCD
  • SHOX2
  • SLC1A3
  • SLC38A1
  • SOBP
  • SOX11
  • SOX9
  • STEAP2
  • SULF1
  • TAC1
  • TBC1D4
  • TBX15
  • TBX3
  • THY1
  • TMEM119
  • TMEM173
  • TNFRSF21
  • TNS1
  • TPBG
  • TWIST1
  • TWIST2
  • WNT5A
  • WNT5B
  • XYLT1
  • ZNF521
  • ZNF618

Down-regulated genes:

  • 1562612_AT
  • 1565809_X_AT
  • 1570469_AT
  • 202350_S_AT
  • 208469_S_AT
  • 209823_X_AT
  • 209826_AT
  • 228560_AT
  • 228750_AT
  • 229968_AT
  • 230537_AT
  • 230828_AT
  • 231239_AT
  • 232562_AT
  • 234034_AT
  • 236586_AT
  • 238307_AT
  • 238718_AT
  • 242270_AT
  • ABLIM2
  • ADAP1
  • ALDH1A1
  • ARHGAP26
  • AVPI1
  • C16ORF82
  • CAB39L
  • CASC10
  • CCDC160
  • CD200
  • CDKL2
  • CLMN
  • CLRN1
  • COL11A2
  • COL14A1
  • COL9A2
  • CRLF1
  • FAM19A5
  • FCGR2A
  • GATM
  • GFRA2
  • GRAMD2
  • HCP5
  • KCND3
  • LCTL
  • LGI1
  • LINC00052
  • LOC729870
  • MAF
  • MARCH1
  • MBP
  • ME2
  • MOXD1
  • MRO
  • MSTN
  • MTL5
  • NEBL
  • PCDH17
  • PPP1R1C
  • QKI
  • RNASE1
  • RP1-249H1.4
  • SESN3
  • SKIDA1
  • SLC15A3
  • SPECC1
  • SPTLC3
  • STOX2
  • TGFA
  • TLR1
  • TMEM139
  • TMEM56
  • TMEM59L
  • TMOD2
  • TRPM3
  • TTYH2
  • UGT8
  • VPS13D
  • VWA1
  • XRCC6BP1
  • ZNF536

In order to find diseases with a phenotype similar to dNFSC we make a direct search (similar expression profiles) in GEO: here you can see the results.

In order to find drugs and compound that might revert dNFSC's phenotype we make an inverse search (opposite expression profiles) in CMap and DrugMatrix: here you can see the results.

Please refer to Section 3.2 if you need help interpreting this results.

5. Known issues.

5.1. Empty tables

When your search returns no results for some table (e.g.: no diseases, no drugs or no interactions), graphics and controls related to those tables will appear blank or showing a message similar to "Could not find column" or "Empty table". These are not errors in NFFinder, just empty results that the dashboard can't show. Figure 12 shows an example of this behavior.

Figure 12. The summary page won't show anything for diseases because NFFinder didn't find any. This is most common when performing searches in CMap/DrugMatrix only, since those databases don't contain disease information.