data_unstranded.bw is sparse and has a different range

[ArnovanHilten]

I loaded the predictions of the model into IGV viewer together with the BAM file and data_unstranded (shown below). For example for this peak:

The tracks show the following:

1. BAM file [0-159]
2. data_unstranded.bw [0-34]
3. cmb chrombpnet_nobias predictions [0 - 3.07]
4. bias model predictions [0 - 0.41]
5. chrombpnet.h5 (cm) predictions [0-27]

Why is the data_unstranded so different from the original BAM profile? The estimated shift was +0/+0. I am also confused by the range and the shape differences between the tracks.

Did something go wrong during preprocessing? None of the predicted tracks is close to the original BAM input.

[ArnovanHilten]

Figured it out, you just keep the starting position of each read.