I recently came across XHMM and got two questions after reading the paper .

How is read depth bias caused by sequence context (probe capture efficiency) normalized? I feel PCA step is not able to do it. Is z-test performed to handle this?

The second question is regarding HMM. I noticed that trio data set is used (Fromer et al. 2012) to establish rate and length of the CNV. Do rate and length here mean transition and emission probability? If so, it is kind of solving Decoding problem (known outcome sequences and model parameters) using viterbi algorithm. Please correct me if I am wrong.

Really appreciate if any one familiar with this tool can share some insights. Thanks in advance.