file *.switch.bed

[cip25]

Hi, I’m reviewing .switch.bed files from haplotype phasing and want to confirm my understanding:

Short: small haplotype switches, likely less reliable.

Long: large consistent haplotype blocks, generally more reliable.

Switchback: regions where the haplotype “switches back,” important to check in triobinning.

When deciding whether to replace or validate a fragment flagged as erroneous or less reliable, should I focus mainly on switchback and long regions, while treating short regions as lower priority?

Any confirmation or reference to documentation/papers would be appreciated.

[arangrhie]

Hi @cip25 ,

Please refer to the Evaluating phasing completeness with hap-mers section and Fig. 4.
The phased_block.bed files contain the assigned haplotype block information, which could guide you to prioritize which one to look at. I use this information along with the hapmer.wig files to determine if there is an unexpected haplotype block in a trio-guided assembly (trio binning or verkko using trio markers etc.).
I left the .switch.bed files for debugging purpose, and for those who are interested to look into the details, but for most cases, the hapmer.wig visualizes where exactly the hapmers are found and should be sufficient for determining what switch / hamming error to investigate.

Note there might be some corner cases where haplotype blocks are expected to switch, such as the immunoglobulin loci in human, where frequent recombination is expected. We discovered this while evaluating the benchmark assembly HG002:

Additionally, while evaluating the assembly for phasing error, we noted large blocks of haplotype switches flagged in the immunoglobulin gene loci. We confirmed this was caused by V(D)J recombination in the LCLs55, making the parental-specific markers unreliable in these regions. V(D)J recombination was confirmed in the NIST reference material by comparison to short reads from a PBMC-derived induced pluripotent stem cell (iPSC) of the same individual (Coriell GM27730). Consequently, all immunoglobulin loci were excluded from parental haplotype switch analyses.

Hope this helps!

Best,
Arang