FastQC 5' end funkiness

[iquigley]

So while waiting for my own data to show up off the hiseq I've been fooling with publicly-available data to make sure I can actually run a pipeline. One thing I've noticed with many datasets from different labs is a curious, and distinctive, pattern of per base sequence content in the first 10-11 bases. You can see it here, here, and another one is attached below. By trimming it in an individual set, though, the % reads aligning doesn't seem to change appreciably. What are these things?

Many thanks for any insights.

[kopi-o]

Start by reading this old thread: http://seqanswers.com/forums/showthread.php?t=14578

[iquigley]

Thank you!