By integrating the three best performing algorithms to infer miRNA mediate regulation from co-expression signatures we have constructed a generalizable Framework for Inference of Regulation by miRNAs (FIRM). FIRM is available as a Python script capable of identifying putative miRNA mediated regulation from transcriptome co-expression signatures.

FIRM was developed in Linux, but by ensuring the following dependencies are met could be modified to work in Windows. In order to function properly this script requires:

1. Python 2.6 or greater (but less than 3.X) - the lastest python can be downloaded here.
2. R - the lastest R can be found here.
3. Weeder - the lastest Weeder can be found here. The paths of this must be configured to identify the location of the frequency files packaged with FIRM. A successful installation of Weeder will allow a user to type 'weederlauncher' at the command line and will run the 'weederlauncher.out' program from Weeder 1.4.2.

We provide FIRM as a tarred and gzipped file:

• Download FIRM.tgz (16.5MB)

    prompt> tar xvzf FIRM.tgz

    prompt> ls FIRM

    FIRM.py  TargetPredictionDatabases  common  exp  miRvestigator.py  pssm.py  weeder_FreqFiles

• FIRM.py - is the script to run the analysis.
• TargetPredictionDatabases - directory containing the target prediction databases PITA and TargetScan.
• common - directory containing general files needed to run the analysis.
• exp - directory where you should put your co-expression signatures files that are to be analyzed. The files should look like this:

      Gene	Group
      NM_000014	32
      NM_000015	23
      ...

  This file is expected to have a header. It has two columns:
  1. RefSeq Transcript ID
  2. Co-expression signature number
• miRvestigator.py - python module of the miRvestiagtor analysis script.
• pssm.py - python module to be a container for PSSM objects.
• weeder_FreqFiles - the Weeder frequence files that should be placed in the appropriate FreqFile location as determined by the way Weeder was installed.

If the dependencies above are met then to run FIRM will simply require the user to create the appropriate co-expression signature files and place them in the 'exp' directory. Then the analysis can be started by typing:

    prompt> python FIRM.py

FIRM limits the Weeder-miRvestigator method to only those inferences of miRNA mediated regulation with a perfect 7- or 8-mer miRvestigator complementarity p-value (p-value = 6.1 x 10^-5 or 1.5 x 10^-5, respectively) to a miRNA seed in miRBase. Inferences of miRNA mediated regulation from the PITA and TargetScan enrichment of predicted miRNA target genes methods were filtered to include only those with Benjamini-Hochberg FDR = 0.001 and at least 10% of genes had a predicted miRNA binding site. After FIRM finished running it will produce a file 'combinedResults.csv' in the main FIRM directory. For the This file will contain a listing of all co-expression signatures that were predicted to be regualted by a miRNA. The column headings are:

1. Dataset - the dataset where the co-expression signature was observed.
2. signature - nubmer for the co-expression signature.
3. miRvestigator.miRNA - miRvestigator predicted miRNA(s).
4. miRvestigator.model - model that fit the miRNA to the PSSM found through Weeder, either 7mer or 8mer.
5. miRvestigator.mature_seq_ids - the miRBase mature sequence IDs for the predicted miRNAs.
6. PITA.miRNA - PITA predicted mirna(s)
7. PITA.percent_targets - percent of co-expression cluster genes with predicted target sites in PITA.
8. PITA.P_Value - p-value for the hypergeometric enrichment of miRNA(s).
9. PITA.mature_seq_ids - the miRBase mature sequence IDs for the predicted miRNAs.
10. TargetScan.miRNA - TargetScan predicted mirna(s)
11. TargetScan.percent_targets - percent of co-expression cluster genes with predicted target sites in TargetScan.
12. TargetScan.P_Value - p-value for the hypergeometric enrichment of miRNA(s).
13. TargetScan.mature_seq_ids - the miRBase mature sequence IDs for the predicted miRNAs.