Table 3 Strengths and limitations of core analysis methods implemented into the pipelines
From: RATEmiRs: the rat atlas of tissue-specific and enriched miRNAs database
Core Analysis Methods Implemented | Strengths | Limitations |
|---|---|---|
NMF | Factors are interpretable | Factorization (W,H matrices) is not always unique |
Reduces dimensions of the data | No statistical inference | |
Fast computation | Convergence can be slow | |
Quasi-Poisson | Has underlying statistical inference | Model dependency and complexity |
Computational simplicity | No probability distribution or log-likelihood | |
Accounts for over-dispersion of the data | Supported by asymptotic (large sample) theory in special cases | |
Requires normalization and transformation of the data | ||
% Total Mapped Reads | Proportion basis offers an intuitive relationship to relative expression | Requires normalization of the data |
Easy to implement | No statistical inference |