Hi! I need advice on the processing steps for my project. It would be really nice to get some feedback on this. I want to perform Trinity de novo assembly from metatranscriptome samples. Since there are different microorganisms inside, I want to use the big assembly as a reference to map then the individual samples (replicates under 2 different conditions) and count the transcripts for testing differential expression. The problem that I am struggling with is: for the big assembly I will have around or more than 200 million reads (PE), depending on how I will process the sequences (I could have 2 big assemblys, each one for the different conditions, and in this case will be less data; or I could maybe get a better "reference assembly" by using all data together) . So I don't now if it will be possible to perform this without problem on the requested resources using Trinity in a HPC cluster, until now I have only used 40 million as a maximum for assembly and its really difficult to keep running a job for so long. Maybe you could give me some advice on how could I improve the data (pre)processing steps?
Another thing that I'm not sure about is: as I dont have a reference genome and I'm not expecting to have a big percentage of further annotation, It would be better for me to use merged PE (longer reads) that its about 20-30% of my sequences, but in this case I would loose the rest of the information from the unpaired reads AND I should treat my sequences in single mode with Trinity... Is there a way that I could combine my merged data with the unmerged and include everything in my analysis? without having to treat everything as single end data? Thanks in advance!