JillAshey
commented
3 months ago Hmm interesting...I'm looking through the Axtell 2013 ShortStack paper and here's their workflow:
And here are my notes from that paper when we first looked at it: Workflow for short stack
- Phase 1: Islands of significant alignment coverage are IDed (islands are defined as continuous genomic coordinates where the depth of alignment coverage exceeds a user-set minimum threshold option –mindepth). Initial islands are padded upstream and downstream by a user-set number of nucleotides (option –pad). Regions that overlap after padding are then merged to form small RNA clusters
- Phase 2: Filters small RNAs of interest based on user-set size range (default 20-24 nt). Loci where the number of RNAs within the user-set size range meets a minimum threshold (default = 0.8) are annotated as ‘Dicer-derived’ and others are annotated as not Dicer-derived (Dicer is an enzyme that cleaves pre-miRNA into miRNA. Unsure of what it means to say a small RNA is ‘Dicer-derived’). The predominant small RNA size of each cluster is reported as the ‘DicerCall’
- Phase 3: Retrieval of genomic sequences for RNA folding
- Phase 4: Identify qualifying hairpin structures. Loci unlikely to be hairpin or miRNA associated are excluded from analysis, as well as clusters annotated as non-Dicer derived and highly repetitive clusters. Potential smRNA-associated hairpins are filtered to retain only those that meet a minimal structural criteria. ShortStack has two preset “miRTypes” for the animal and plant kingdoms, which increase specificity by excluding structures with unusual characteristics (what qualifies as an unusual characteristic?). Animal miRType excludes 15% of all annotated miRBase structures for metazoans and 6% of all known viral miRBase structures. Plant miRType would exclude 12% of all known miRBase structures from Viridiplantae
- Phase 5: Annotation of hairpin-associated and miRNA. Annotation requires structural and expression-based criteria. Annotation of a locus as a miRNA indicates there is sufficient evidence to justify annotation without assistance from prior knowledge (ie prior annotation as a miRNA for conservation in another species). Hairpins with strong evidence of causality with respect to observed small RNA patterns are annotated as either hairpin-derived (HP) or miRNA loci
- Phase 6: Reports produced including summary file, alignment details for HP or miRNA clusters and a .bed formatted file
My guess is that something weird is going on in Phase 4 or 5? I'll continue looking into it.
I was also looking at the Shortstack code using the merged reads w/ the cnidarian miRNA database. When I look at the venn diagrams in the Apul code and the Pmea code, the database miRNA numbers don't seem to match up. In Apul, the database miRNAs are 68 and in Pmea, the database miRNAs are 29. The venn diagram for the Peve code has 36 database miRNAs. Is there some difference between the databases that are being used? Or am I just bad at counting?
kubu4
shedurkin
sr320
I was taking a look at database matches among the three species and noticed something odd. All three species have miRNAs with database matches to publications of mir-2030, but that was not one of the miRNAs identified as conserved across the three species. I compared the mature miRNA sequences for these loci and, while the Apul mir-2030 and Peve mir-2030 mature sequences were identical, the Pmea mir-2030 mature sequence was very different. How, then, did all three of these sequences have database matches to the same known miRNA (mir-2030)?
I finally realized that the Pmea mir-2030 star sequence was identical to the Apul and Peve mir-2030 mature sequences. This raises the question -- how confident can we be that ShortStack is correctly distinguishing the mature and star sequences? I'm not actually sure how it makes this distinction, and I haven't had the chance yet to search for other instances of this happening.
If anyone wants to look into this, the links to shortstack outputs are below Apul: https://github.com/urol-e5/deep-dive/blob/main/D-Apul/output/13.2.1-Apul-sRNAseq-ShortStack-31bp-fastp-merged-cnidarian_miRBase/ShortStack_out Peve: https://github.com/urol-e5/deep-dive/blob/main/E-Peve/output/08.2-Peve-sRNAseq-ShortStack-31bp-fastp-merged/ShortStack_out Pmea: https://github.com/urol-e5/deep-dive/blob/main/F-Pmea/output/13.2.1-Pmea-sRNAseq-ShortStack-31bp-fastp-merged-cnidarian_miRBase/ShortStack_out