Unevenness of sequencing throughput in targeted capture sequencing

Question

[arthurhk]

Hi All Recently, we have sequenced around ~190 tumor samples with our customized targeted capture panel. We used the capture probes ordered from IDT. During capture, we included ~20 samples together in a capture pool, for each library, we added 250ng library DNA for capture (the library concentration was measured by qubit). After sequencing, we found that the sequencing throughput was quite uneven, some samples has around 2 to 3-fold differences in sequencing depth (the raw sequencing depth is range from 1000X to 3000X). Can I ask for help for the following questions? 1. Is this unevenness commonly observed in targeted capture sequencing? 2. Theoretically, the higher the depth, the more variants can be identified, will this unevenness in sequencing depth have a big impact during data comparison across the same batch of samples? 3. If yes, is there any method to normalize the data to reduce the impact? Thank you very much, Arthur

[arthurhk]

Thx a lot.

Arthur

Answer 1

[ATpoint]

1) Uneven depth can be the result of many factors, starting from improper individual library quantification over differences in capturing efficiency (e.g. purity differences between DNA samples). The first one probably being the most important aspect here.

2) Depends how the variant caller works. You can always subsample high-depth samples to match the average depth across your pool.

3) Yes, e.g. Selecting Random Pairs From Fastq?. The keyword is subsample fastq, please use the search function, multiple threads on this but the seqtk solution in the one I linked should work fine, alternatively, simply take the first n reads from each fastq file since the order is actually already random when it comes from the sequencer.