transform ANI matrix to 1-D condensed distance matrix

[danpal96]

ANIclustermap v2.0.1

I was using ANIclustermap to compare some genomes and found some discrepancies between what the visualization said were the more similar genomes and which were according to the ANI matrix.

After some digging, I found that when in the next line you calculate the linkage:

ANIclustermap/src/aniclustermap/scripts/aniclustermap.py

Line 175 in afed77a

linkage = hc.linkage(ani_matrix_df, method="average")

According to the documentation since you gave it a 2D matrix it will calculate the euclidean distances between the rows (and columns). I think that a better approach is to transform the ANI matrix to a 1-D condensed distance matrix so it can use it directly as the distance matrix using the squareform function:

linkage = hc.linkage(squareform(100 - ani_matrix_df), method="average")

ANI matrix using the example files and skani mode:

	GCF_000009725.1_ASM972v1_genomic	GCF_000270265.1_ASM27026v1_genomic	GCF_000478825.2_ASM47882v2_genomic	GCF_018282115.1_ASM1828211v1_genomic	GCF_018845095.1_ASM1884509v1_genomic	GCF_019924095.1_ASM1992409v1_genomic
GCF_000009725.1_ASM972v1_genomic	100	100	86.41	88.12	99.99	88.09
GCF_000270265.1_ASM27026v1_genomic	100	100	86.46	88.57	100	88.58
GCF_000478825.2_ASM47882v2_genomic	86.41	86.46	100	84.83	86.41	85.33
GCF_018282115.1_ASM1828211v1_genomic	88.12	88.57	84.83	100	88.13	99.46
GCF_018845095.1_ASM1884509v1_genomic	99.99	100	86.41	88.13	100	88.09
GCF_019924095.1_ASM1992409v1_genomic	88.09	88.58	85.33	99.46	88.09	100

Heatmap with current behaviour:

Heatmap with 1-D condensed distance matrix:

GCF_000478825.2_ASM47882v2_genomic is consistently the genome with lower ANI to all other genomes and now this is also reflected in the heatmap with the 1-D condensed distance matrix.

I will make a PR implementing this.