In terms of how successful your assembly was, do you have basic stats such as the N50, number of contigs, distribution of length of contigs?

Assuming this is not a model species, do you have any idea what the gene content should be like in the plant your work with (there are some well annotated plants to compare with, via Blast etc.)?

You can use a tool such as CEGMA/BUSCO to assess the "completeness" of your assembly. I believe this will tell you something about what proportion of core eukaryotic genes are present in your assembly as well as the number of partial vs. complete transcripts.

These are just a few things that come to mind, I'm sure there are several other options.

Thank you Cameron. We got basic statistics of the assembly. We tried TRINITY with strand specific and non stand specific options. We doubt the assembly because the out put trinity.fasta we got is of very small size viz; two biological replicates after quality trimming were of 3.4GB and 3.1GB, which generated 1.7MB trinity reference.  Let me try with CEGMA/BUSCO.

The best way would be try with a few other assemblers like SoapDenovo or Velvet/Oasis. Find the best transcript via EvidentialGene (http://arthropods.eugenes.org/EvidentialGene/)

What you can do to extract the strand information is to quantify how large a fraction of reads map to you predicted transcripts, and large a fraction maps to the reverse complement (simply switch the strand information to the opposite strand).

If you get ~50/50 it is unstranded and if you get >75/

Thank you Loke, I was planning to compare my trinity out put with other assemblers too.

Thanks Kristoffer, the information is really valuable. Let me check with all the details provided. Do cufflink work good with non model plants?

In theory cufflinks should work fine since it is a probabilistic framework of a RNA-seq experiment. But how it works in practise I do not know. One would have to spend some time optimizing the many parameters (especially allowed/expected intron-lengths).

With regards to optimizing the Baysembler migth be easier since it have fewer paramters (and it migth also perform better: http://www.ncbi.nlm.nih.gov/pubmed/25367074 )

Thank you . I think its better to compare my results with other assemblers also.Though i could run trinity,I am totally confused about the size of trinity.fasta output and lower contigs.

I agree that you should try some other assemblers and compare results. I tried Bridger recently and believe it works quite well, it's also splice-aware which can help if your transcriptome has lots of splice isoforms. SOAPdenovo trans is another good option.

You mention that your Trinity output file is only 1.7Mb. How many transcripts were assembled and what was the length distribution like? It might help if you post some of the assembly metrics to give everyone an idea of what the problem could be.

Have you taken a look at any of the reads that did not get assembled? You could just Blast a few and it might give an indication if there was some major contamination in your sample, eg. from other organisms or from the sequencing run itself. Might not be very likely, but doesn't take long to check.

 Thank you cameron, The total contigs from trinity.fasta file was 3257. I can attach the screenshots of  basic statistics and contigs with the trinity assembly.I don't know what are the steps i should do to get data like number of scaffolds, N50 count scaffolds, lenghth distribution, min length of contig, etc. Do i have to run it on linux using some perl script?

I use trinityStats and CEGMA to get statistics and get completeness of the transcriptome, as suggested above. Anyway, if you are assembling it de-novo I would suggest that the best way to check the "quality" of the assembly is to get the annotation of the transcripts and look at the number of results and statistics about the alignments. Then you can try to do the same with another assembling software.

Thank you. I used TrinityStats.pl for the basic statistics.I think i should try more.

Also while cross checking TRINITY output, i could found one file named single,fa (1.7GB). I think it could be the unassembled sequences. Am i  right or not?

Judging by the stats in the screenshot you posted this assembly seems far from complete. You have very few genes assembled (2253) relative to the typical gene content of plants (>20000). Also, your transcripts/contigs seem to be very short, with an average length of the 'longest isoform for each gene' being only ~300nt. I think you're right to be wary of this assebmly.

This is single end Ion Torrent data? What does the QC of your original/cleaned reads look like? I'm more familiar with Illumina data, so I'm not sure what types of sequencing-based problems can happen with Ion (eg. primer-dimers).

I haven't used Trinity in a while so I'm not too sure about "single.fa" but I think it's some kind of intermediate file and it might be unused reads as you say. Did you tell Trinity to use your input as single-end reads? I think Trinity expects paired-end data by default.

QC of the original clean reads looks good with NGS Fast QC.Sure, i have specified in the trinity command  the reads were single end. Like you said, several reports also shown that TRINITY works good with paired end illumina RNA seq data. So I am not planning to stick to trinity alone. 

Thanks a lot cameron for your suggestions.