Introduction to Bioinformatics and Computational Genomics

Week 4Sequencing technologies and genome assembly

How reads are made and what they cost you in quality, then the two graph algorithms that put them back together, and the metrics that tell you whether it worked.

Questions this week answers

  • Why does a repeat break an assembly, and what actually fixes it?
  • Two assemblies both report an N50 of 4 Mb. What might still be completely different about them?
  • Q20 is one error in a hundred, which sounds harmless. Why does anyone trim?

By the end of this week you can

  • Read a Phred score as an error probability
  • Explain why overlap-layout-consensus does not scale and de Bruijn graphs do
  • Compute N50 and coverage from an assembly
  • Say what N50 does not tell you
0 of 7 done

  1. What N50 actually measures

    Drag the contigs. Watch how far the N50 can drift from the mean without a single base being added or removed.

    Half of these 18.5 Mb sit in contigs of 4 Mb or longer. It takes 2 of the 7 contigs to get there.

    N504 Mb2 contigs to reach half
    Mean length2.64 Mb7 contigs
    Median length2 Mb
    Assembly total18.5 Mbhalf is 9.25 Mb
    Contig lengths7 contigs
    6.0 Mb
    4.0 Mb
    3.0 Mb
    2.0 Mb
    1.5 Mb
    1.0 Mb
    1.0 Mb
    Drag a bar to resize it. Add or remove contigs to change the assembly.
    Compare against an expected genome size
    N50 only compares fairly between assemblies of similar size. NG50 fixes that.
    The size the genome is believed to be, independent of what you assembled.
    Things to try0 of 3

    Default lengths from the week 4 assembly lecture.

What should survive this week

  • Finding a Hamiltonian path is NP-complete and finding an Eulerian path is linear. That single fact is why de Bruijn graphs displaced overlap-layout-consensus.
  • Repeats split an assembly because the information needed to order the pieces is genuinely not in the reads. It is not a failure of the assembler.
  • N50 is a length, not a count, it is deaf to the short end of an assembly, and it is only comparable between genomes of similar size. NG50 exists to fix the last part.
  • Per-base quality and per-read quality are different questions. At Q20 only 22% of 150 bp reads are error-free.