MiRNAss is the first semi-supervised method for microRNA prediction. It is specifically designed to face the problem of scarce pre-miRNAs that can be used as positive examples, in the context of a real prediction task from genome-wide data.
This web-demo allows to test miRNAss method under different conditions and parameters with small datasets. The input parameters are:
- Features file: comma separate value file with the features extracted from hairpin sequences.
- P: number of real pre-miRNAs to use as training samples (P>0).
- N: number of non pre-miRNAs to use as training samples.
- k: neighbours used to build the nearest neighbour graph.
- Scaling: apply RELIEF for features scaling.
To test the algorithm with a few number of positive training samples, use the provided sample file (small Human dataset), P in [1…128], N=0, k=10 and no scaling.
Contact: Cristian Yones