ANI.dendrogram hclust error

[jessih26]

Hello,

Thank you so much for developing and sharing this tool! I am trying to make a tree from fastANI pairwise data. After successfully importing the txt file, I received the following error when I ran dend <- ANI.dendrogram(bactaxRObject = ani, ANI_threshold = 95) on my dataset:

Error in hclust(d = d.dist, method = "average") : NaN dissimilarity value.

I double-checked that my data dose not contain any NAs by running

is.na(ani@ANI)%>%sum() [1] 0

I am just wondering if you have any insight on how to fix this and it seems like that the hclust command introduces some NAs. Thank you.

Best, Jessi

[lmc297]

Hi Jessi,

It seems there might be an issue when your ANI values are being converted into a dissimilarity matrix (which is almost certainly not your fault!). It's hard for me to know exactly what's going on without seeing your data, but I would try the following (where ani is your bactaxR object, per the code you have above), and replying to this comment with the output you get:

1) Just to double-check everything loaded correctly: summary(ani)

2) Just to get a quick summary of your ANI values to make sure there's nothing out of the ordinary/weird: summary(ani@ANI)

3) Another quick test, i.e., a histogram, just to be safe: ANI.histogram(bactaxRObject = ani)

4) Try copying and pasting the code chunk below into RStudio and running it line-by-line; in addition to figuring out if/where any errors are occurring, I have some sanity checks in there to see if there are any places where things are not going as planned :

# convert your bactaxR object to a data frame
fastani <- data.frame(ani@query, ani@reference, ani@ANI)

# add column names to the data frame
colnames(fastani) <- c("query", "reference", "ANI")

# convert to a square matrix, add rownames, drop query column
s <- dcast(fastani, formula <- query~reference, value.var = "ANI")
rownames(s) <- s$query
s <- as.matrix(s[ , !(colnames(s) == "query")])

# as a sanity check, let's look at the dimensions of our matrix
dim(fastani)
dim(s) 
# s should be square, where nrow(s)*ncol(s) should equal nrow(fastani)

# another sanity check: check that the values of s make sense, i.e., no NAs
summary(as.vector(s))

# create a square matrix of 100s
j <- matrix(data = 100, nrow = nrow(s), ncol = ncol(s))

# convert s to a matrix of dissimilarities d by subtracting from j
d <- j - s

# sanity check: dimensions of d should equal dimensions of s
dim(s)
dim(d)

# sanity check: dissimilarities make sense and there are no NAs
summary(as.vector(d))

# make the dissimilarities symmetric
d.sym <- 0.5 * (d + t(d))

# again, a sanity check: dimensions of d.sym should equal those of d and s
dim(d.sym)

# another sanity check: d.sym values make sense, are not NA
summary(as.vector(d.sym))

# convert d.sym to a distance object
d.dist <- as.dist(d.sym)

# sanity check: make sure d.dist doesn't have NAs
summary(as.vector(d.dist))

# run hclust on d.dist and plot, see if any errors appear
h <- hclust(d = d.dist, method = "average")
plot(h)

Hopefully this might help identify where the error(s) occur! I'd also be happy to look at your raw data, if you prefer.

Let me know if this helps,

Laura

[jessih26]

Hello Laura,

Thank you so much for your thorough explanation! I found that somehow fastANI did not perform pairwise comparison across all possible pairs so my s matrix is not a perfect square. I switched to my AAI results generated using other tools and successfully got the dendrogram. Thanks a lot for your help!

Best, Jessi