De novo assembly for RNA-seq Illumina PE reads

Question

[mcff23]

Hi everyone!

I'm new with RNA-seq analysis..

Wich would be the best tool to do a de novo assembly of RNA-seq Illumina PE reads (2x100 bp) with a 32 Gb RAM server?

Thanks in advance!

MF

[st.ph.n]

There's plenty out there: Trinity, Oases, Trans-Abyss, Cufflinks; to name a few. It depends what your end goal is. How many reads in your files? I don't have much experience with the others, but I know for Trinity, you can adjust the memory usage on the command line. I ran a SE assembly that had approx. 60 million reads on a workstation with 12Gb RAM.

Answer 1

[mcff23]

Hi Stephen!

Thank you so much for your quick answer!

For Trinity I found a Jellyfish memory parameter which I was already using, I couldn't find any other for memory adjustment.

Is there another one for this purpose? I have 196 Mi PE reads.

Thanks!

MF

[st.ph.n]

check all paramters by only using the --show_full_usage parameter. You can specify butterfly usage. Also, as Matt suggests, use the in silico normalization parameter which will reduce your number of reads given a sequencing depth coverage threshold (default 50x). Also, make sure you QC your reads: trim barcodes/adapters left over from sequencing, and low quality bases. You can use the trimmomatic parameter to trim your reads if you don't have a program of choice. Sickle is also a good one if you want to run it prior to Trinity.

Answer 2

[Matt]

For Trinity, the only way to adjust is using the -JM setting, which you have already discovered. This only effects memory usage in 1 stage of the assembly, so not exactly sufficient. Depending on the number of reads you have, you may want to use digital normalization. You could also increase the --min_kmer_cov to 2, but this will almost always result in a more fragmented assembly.

[mcff23]

Hi Matt! I have 196 Mi PE reads, do you suggest me to normalize?