Do not trim "head" of the reads .. unless these are known to be some sort of inline barcodes.
@James has already referred to the possible cause where we see this pattern. See this post for additional information.

If you're referring to the per base sequence content of the first 10 bases. This looks like insertion bias from the Nextera kit used to prepare sequencing libraries. Do you know what kit was used for sequencing? If it was a Nextera kit, you only need to worry about removing Nextera adapters.