CLIP-based scoring

Instead of scoring a sequence against a PWM, -clip scores it against real binding evidence — peaks from a CLIP-family experiment (eCLIP, iCLIP, etc.) supplied as a BED/bedGraph file:

scanRBP loci.fasta -clip peaks.bed.gz -protein MY_RBP -heatmap clip_scores

This is the alternative to PWM-based scanning: rather than asking "does this sequence look like this protein's known motif," it asks "did this protein actually bind here, according to this experiment." -protein here just labels the output row/track — it isn't used to look up a database entry, since the signal comes entirely from peaks.bed.gz.

Sequence ID format

CLIP-based scoring only works with a FASTA file (not a single command-line sequence), and each FASTA sequence id must encode its genomic locus so scanRBP knows where to look up peak values:

<chromosome><strand>_<start>_<stop>

For example, a FASTA record for chromosome 1, plus strand, from position 11012300 to 11012400:

>1+_11012300_11012400
ACGTACGT...

The strand character (+ or -) is the last character before the first underscore-delimited coordinate; everything before it is the chromosome name (chromosome names containing underscores are handled correctly, since only the last two _-separated fields are treated as start/stop).

How it works

scanRBP loads the peak file once with pybio.data.Bedgraph, then for each sequence, calls get_vector(chr, strand, start, stop) to pull the per-position peak values for that exact locus — so start/stop in the sequence id must match real genomic coordinates in peaks.bed.gz's coordinate system. -nonzero still applies (clip negative values to 0), and output files follow the same .tab.gz / .png / .pdf convention as PWM scoring — see Output files.

See Motif database & search for a pointer to ENCODE's public eCLIP experiments if you don't have your own CLIP data.