How can you check if the downsampler in indeed random and efficient

Question

[shruti]

Hi, I am using the Picard downsampler for downsampling around 50 million reads from the entire human transciptome for my experiment. I want to know if the downsampler is indeed random and not creating bias. Is there anyway I can check that ?

-Shruti

[GouthamAtla]

Downsampling reads from a bam file using DownsampleSam?

[shruti]

Yes, i've used tophat for generating the BAM files from which I have used DownsampleSam from Picard.

Answer 1

[GouthamAtla]

One way is to validate the results by generating different random sets and compare them wrt your goal. For e.g. If you have annotations, you can do quantification of genes, and check if they correlate between different random sets to get an idea.

Otherwise, if you go into details, it uses java.util.Random, which depends on ["Linear congruential generator"] algorithm to generate random numbers.

[shruti]

Thank you @Goutham!