We have been using Oases for so long that we failed to report the ‘official’ announcement of its arrival.
On comparison between Oases, Trinity and transAbyss, a figure from the above paper has all information you need (Hint. ‘Oases is the best’)-
On comparing de novo and reference-based assemblers (i.e. Cufflinks) -
As could be expected, Cufflinks generally outperforms the de novo assembly algorithms, as it benefits from using the reference genome to anchor its assemblies (Fig. 3). Nonetheless, it is interesting to note that as expression level and therefore coverage depth go up, the gap narrows.
On finding alternatively spliced regions -
For each assembled transcript, the average number of additionally assembled transcripts from the same gene are, respectively, 1.21, 1.25, 1.01 and 1.56 for Oases, transABySS, Trinity and Cufflinks. Cufflinks performs better in that respect, whereas Trinity is less sensitive.
Few personal comments -
1. We did a run time comparison of Velvet+Oases and Trinity that the readers may check. We did not compare outputs.
The readers may also find this commentary helpful in understanding how different programs work.
2. Two questions may pop up in your mind - (i) Does Oases really give significantly better transcriptome than Trinity?
(ii) Is there anything in Oases algorithm that allows it to produce better transcriptome than Trinity?
As per the output presented in above comparative figure, both Velvet and Oases assembled 822 genes, whereas Trinity identified 180 additional genes and Oases identified 263 additional genes. So, Oases did not find 180 genes that Trinity did and Trinity did not find 263 genes that Oases did. Trans-Abyss had another 222 that neither Velvet nor Trinity identified. So, most important conclusion from the figure is that the set assembled by transAbyss+Oases+Trinity is significantly bigger than what each program assembled alone.
One thing to note is that Oases paper had the advantage of coming in last and seeing and improving upon other methods. For example, Trinity has fixed k-mer size of 25, whereas the Oases trial was run with variable k-mers (19-35nt). We do not know how much the Trinity results would improve with longer k-mers.
In these experiments, Oases tends to be more sensitive, Trinity more accurate. The correlation of small k-mer assemblies and misassembly rates suggests that homologies between genes are the main source of errors. As reads get longer, and coverage depths greater, sensitivity will only increase and users will probably avoid the shorter k-mer lengths for greater accuracy. Short k-mers will only be necessary to retrieve the very rare transcripts.