In this project, we have devised a RNA-Seq differential expression pipeline to investigate whether mapping to a reference genome could be circumvented while still mainting accuracy. To make the results comparable, we deviated from Sleuth analysis and used DESeq instead for fold change calculation.
We have shown that ,generally, the different methods provided very similar outcome (significant overlap of differential expressed genes, correlated estimated abundances), especially between the mapping-free methods, mainly because of the underlying algorithmic similarity. Although all methods captured the basic biology of the experiment (overexpression of immunorelated genes), discrepancies at lower fold changes were observed but their importance could not be verified due to the absence of a benchmark dataset- although inspection of some of them revealed different isoform transcripts that could have accounted for the observed differences.
Pertaining to the choice of best method for differential expression analysis, one should consider the characteristics of the experiment -for example presence of different isoform abundances should point towards the mapping free methods. Overall, our project showed that mapping-free methods could be a robust and fast alternative to the normal pipeline, although the limited sample size and the absence of a benchmark should be considered as main limitations of our study to generalize the noticed trend.
We have shown that ,generally, the different methods provided very similar outcome (significant overlap of differential expressed genes, correlated estimated abundances), especially between the mapping-free methods, mainly because of the underlying algorithmic similarity. Although all methods captured the basic biology of the experiment (overexpression of immunorelated genes), discrepancies at lower fold changes were observed but their importance could not be verified due to the absence of a benchmark dataset- although inspection of some of them revealed different isoform transcripts that could have accounted for the observed differences.
Pertaining to the choice of best method for differential expression analysis, one should consider the characteristics of the experiment -for example presence of different isoform abundances should point towards the mapping free methods. Overall, our project showed that mapping-free methods could be a robust and fast alternative to the normal pipeline, although the limited sample size and the absence of a benchmark should be considered as main limitations of our study to generalize the noticed trend.
Δεν υπάρχουν σχόλια:
Δημοσίευση σχολίου