limey-bean / Anacapa

Written by Emily Curd (eecurd@g.ucla.edu), Jesse Gomer (jessegomer@gmail.com), Gaurav Kandlikar (gkandlikar@ucla.edu), Zack Gold (zjgold@ucla.edu), Max Ogden (max@maxogden.com), Lenore Pipes (lpipes@berkeley.edu)and Baochen Shi (biosbc@gmail.com). Assistance was provided by Rachel Meyer (rsmeyer@ucla.edu).
MIT License
42 stars 19 forks source link

Figures - to do #42

Closed gauravsk closed 6 years ago

gauravsk commented 6 years ago

Crux vs. the world

Bowtie2-BLCA v Blast-BLCA v Qiime2

All comparisons are happening at the 60% cutoff

Hi-Seq vs Mi-Seq

Anacapa vs. the world

gauravsk commented 6 years ago

@limey-bean Can you check this call: KU900226.1_extraction 1 4 0 2 in the file Anacapa/data-for-figs/Database_comparisons/compare-18S-databases/18S_V8-9/CRUX_unfiltered_18S_V8-9_coded/coded_60_taxonomy.txt? The higher blca cutoff files (70 to 90) have 5 "corrects" in place of the 4 here, which doesn't make sense.

EDIT: I made a copy of the heatmap and barplot assuming this to be true. They're in the commit https://github.com/limey-bean/Anacapa/commit/1e6b3cf7a246f6089472bcbcfa6f615e4c0a2e32 . Let me know if these should be the final ones.