miRTPred predicts for an miRNA and mRNA. The overall execution workflow of miRTPred is given in the 'about' tab. The following program generates the following for an miRNA & mRNA pair.
- Target context features
- The probability scores for each of the potential target binding site.
In this example run, we use hsa-let-7b-5p and NM_000016 as the probable target interaction pair. This program as of now, accepts RefSeq id for transcript. The following steps are required to be followed in order to generate predictions.
Prerequisites
miRTPred execution requires Vienna RNA package 2.0 or later, scikit-learn python package, mlxtend package, miRanda and bedops. The miRTPred executable can be downlaoded from the download tab of the site. Run all the scripts from miRTPred_v0 directory. Give executable permission to the executable program as required.
$ tar -xzvf miRTPred_v0.tar.gz
$ cd miRTPred_v0/
Annotation
All annotation files as well as the model file required to run the program are included in the miRTPred archive file.
Step I: Conservation Score Generation
Run the provided script. The script takes input the refseq id of the transcript
$ sh gen_cons_score/gen_cons.sh NM_000016
Step II: Feature Generation
Make sure to give executable permission using chmod +x gen_feats/gen_features
$ sh sequences/processing_sequences.sh hsa-let-7b-5p NM_000016
$ ./gen_feats/gen_features sequences/hsa_mature.fa sequences/3utr_hsa_final.fa gen_cons_score/NM_000016_3utr.format.phast sequences/pairs.txt > gen_feats/features.out
$ sh annotations/run_proc.sh NM_000016
Step III: Run Prediction
Make sure to have python, scikit-learn packages and mlxtend packages installed before executing this.
$ python predict/run_model_evaluation.py -t gen_feats/select_features.out.csv -m predict/optimal_hard_voting_calib.pkl -o tt.csv