identify Y contigs

[ptrebert]

implement a simple strategy (i.e. rule-based) to identify chrY contigs in the de novo assemblies.

Current rule set:

• [x] 1. tig has only chrY alignments
• [x] 2. tig has mixed alignments but Y-specific motif hits (above threshold)
• [x] 3. tig has primary alignments to chrY with many (current T: >300) unspecific motif hits (above threshold)
• [x] 4. tig is unaligned and has Y-specific motif hits (above threshold)
• [x] 5. tig has more than 90% bp primary alignments to chrY

Motif hits above threshold = "high-quality hits", thresholds set by expert curation

[ptrebert]

First observations: so far, rule 1 captures something like 90+ percent of all contigs (i.e., not necessarily assembled Y bp), and rule 2 triggers occasionally.

For 18989, the beginning (~PAR1 up to CEN) is missing, and it looks like there are no motif matches for the contig, and some spurious alignments to other chromosomes. Hence, rule 5 should also be implemented with a stringent threshold (say, >90% primary alignments to chrY?) @pilleh what do you think?

[pilleh]

For the Y short arm - kind of makes sense that it might be missing (PAR1 and ~3Mb of XTR probably can map to chrX), but I would expect most of the primary alignments to be to chrY, at least from XTR. not sure what the best threshold would be - could try >90% and see what happens? I'm kind of expecting most of the Yp to be assembled as one large contig, which hopefully simplifies its identification, but PAR1 might be more fragmented.

[ptrebert]

for 18989, there is a contig spanning

AMPL1
AMPL2
HET1_centro
other1
PAR1
XDR1
XDR2
XTR1
XTR2

that is missed with the current selection strategy

try >90% and see what happens?

maybe another "thresholding" problem to revisit when all data are in and we can look at the final distribution; I go with >90% for now

[pilleh]

For 18989 - yep, looks like the whole short arm then.

[pilleh]

@ptrebert In addition to the steps in the pipeline - we can further confirm the origin of the Y contigs using HiC, and probably BioNano. So all in all, our approach to identify Y contigs should be pretty waterproof.

[ptrebert]

rule 5 has been implemented, and for 18989, the missing contig is now selected (just testing locally, this change is not yet propagated throughout all results).

Re HiC/Bionano: yes, let's see how that goes.

done c05672c