3.1 Data statistics following import

These are harvested from the demux summarize file

per.sample.count <- read.csv("per-sample-fastq-counts.csv")
datatable(per.sample.count)

4.1 results

qfil <- read.csv("qtrim.stats.csv", stringsAsFactors = F)
qfil <- qfil[-1,]
qfil$total.input.reads <- as.numeric(qfil$total.input.reads)
qfil$total.retained.reads <- as.numeric(qfil$total.retained.reads)
datatable(qfil)

Total read count has definitely dropped, by about 50% in most cases.

print(paste( sum(qfil$total.retained.reads)/sum(qfil$total.input.reads) ,"of the reads were retained after quality trimming"))

## [1] "0.501296204281795 of the reads were retained after quality trimming"

There are several steps in the background, see the script for details.

4.2 chimera removal

chim <- read.csv("per-sample-non-chim.csv", stringsAsFactors = F, header = F)
colnames(chim) <- c("sample.id", "non.chimeric")
chim$`non.chimeric` <- as.numeric(chim$`non.chimeric`)
chim <- left_join(qfil,chim)
chim <- chim[-c(4,5,6)]
datatable(chim)

print(paste( sum(chim$non.chimeric)/sum(chim$total.retained.reads) ,"of the reads were retained after chimeara filtering"))

## [1] "0.970027365381715 of the reads were retained after chimeara filtering"

5.1 weighted-UniFrac PCoA ordination

orange=inland, red=coastal_1, blue=coastal_2

5.2 unweighted UniFrac PCoA ordination

In the unweighted UniFrac based ordination, the coastal and inland sites do not separate cleanly, indicating many shared closely-related taxa. Note the weakness of the axes in the unweighted version (says that community structure differences are less clear, much more multi-factorial). When ordinated based on a weighted distance matrix, which takes relative abundances of taxa into account, the sites become more clear; the coastal columns cluster apart from the inland site, unsurprisingly. In the weighted version, two axes explain over 65% of community structure (vs. 21% in unweighted version). However, there is remarkable overlap/similarity along the main axis of variation between inland and some samples in the coastal sites, meaning that there is an imporant “sub-community” of similar taxa at similar relative abundances. I could not determine any physical/chemical parameters that might explain this similarity.

6.1 Observed OTUs

6.2 Shannon Diversity

• at 400 sampling depth, diversity is leveled off decently
• OTU count (richness) is rising drastically
• Inland site is more diverse than coastal sites overall

6.3 OTU-tables

All tables are provied in table format. A broad overview is taken here.

“Phylum level”

The overall difference between inland and coastal is striking, with Firmicutes being the most dominant in inland sites and Proteobacteria dominating coastal. Perhaps explained by water levels and increased salinity.

Difference between the two coastal columns is also obvious, with heavy cyanobacterial influence (pink: this is actually just chloroplast mostly, will have to be screened out?) at site 1 and heavy Euryarchaeota/Gemmatimonades co-dominance at site 2, although this is not a consistent trend.

6.4 Interesting observations

gem.halo <- read.csv("analysis/PAUC43.halobacteria.csv")
library(ggplot2)
ggplot(gem.halo, aes(x=PAUC43f, y=Halobacteria)) + geom_point()

6.5 Additional observations

casual observations on particular taxa, fishing for stories