What are the potential sources of errors in sequencing datasets?

From an e-mail discussion with C. Titus Brown.

I could think of the following:

• PCR amplification errors (with amplicon sequencing, or all Illumina sequencing as there are usually 5-10 cycles of PCR performed before cluster generation):
  • SNPs
  • PCR chimeras (very likely to be a problem in a metagenomic dataset)
  • Indels (sometimes strand-specific)
• (Illumina) Cluster generation errors - probably similar to PCR as this is a PCR-like stage
• (454/PGM) emPCR errors - again similar to PCR
• (All platforms) Random sequencing errors
• Systematic sequencing errors: seen in Illumina in the past; certain GC-rich motifs, inverted repeats, downstream of homopolymers SNPs, 454/PGM: homopolymers
• Adaptor sequencing
• Post-adaptor read through
• Sample contamination
• Genuine biological variation (not sequencing error, but could be confused)

Others??