You can use ggplot2 as recommended by Irsan in this post.

Edit: the link works now.

If you are working with BED files, at a very basic level, you could use bedmap to count the number of features that overlap positions of interest ("density"):

$ bedmap --echo --count positions.bed features.bed > answer.bed

You can then apply whatever model over the desired window size, to compare what you expect with what you observe.

For instance, if you pick windows at random over all positions in your background space and --count features over those windows, that gives you a background expected rate. Then you can compare that with what you observe with the observed density of features in windows at each position.