two-step approach with fast pre-filter

[martin-raden]

Following the results and repeating the benchmark from https://doi.org/10.1093/bib/bby032
we want to test whether a combination of a fast pre-filter tool like guugle will improve IntaRNAs runtime and prediction accuracy.

• run guugle or similar fast method to identify potential interactions
• combine interaction ranges to predictions ranges (eg on target only)
• run intarna on pre-filtered ranges only

[martin-raden]

also check sTarPicker: A Method for Efficient Prediction of Bacterial sRNA Targets Based on a Two-Step Model for Hybridization

guugle-based with seed-extension

[martin-raden]

• ASSA https://doi.org/10.1142/S0219720018400012

[martin-raden]

prefilter query sequences using its accessibility profile (precomputed or via RNAplfold or IntaRNA
- filter matches based on maximal GU/AU content (check statistics on seeds of genomewide screens)
- generate target region list
- evaluate list features (length distribution, (estimated/modelled) time for ED computation etc) and compare to full target set