[mendel]

Number of reads per lane

How many reads one could get from each lane of HiSeq? I am talking about RNA-Seq. Illumina rep told me between 45-60M reads/lane. This is kind of low?

[HESmith]

You can run any number of cycles on the HiSeq. The 200 cycle kit is actually 209 cycles (to allow for multiplexing). By dividing into appropriate aliquots, we've gotten three 36 cycle plus two 50 cycle runs from a single kit (plus one extra!).

For short runs, the cluster density can be increased above the recommended maximum. We've tested with phiX control at 11 pM for 36 cycles, and obtained 100 million reads that passed filter (94% with no errors, and >99% aligned). But your mileage may vary.

[lingfung.tang]

illumina recommends a cluster density of 425K/mm2. but in real life, what would you target to? higher or lower?

[HESmith]

For short reads (36bp), we've pushed it to 700K. At that density, phiX yielded 99 million reads that passed filter, with an error rate of <0.5%. For long reads (paired-end 101), 425K is the optimum.

[lingfung.tang]

Thanks. But any reason of pushing 36bp sequencing to 700K and keep it to 425K for 2x100bp? How optimum is being defined? Thanks.

[HESmith]

Optimum depends upon the application, which also influences the type of sequencing you choose. The error rate increases with each cycle. De novo assembly and SNP mapping work best with longer reads and are less tolerant of sequencing errors, whereas short reads with more errors are generally sufficient for transcript counting.