Fig. 5 - Interpreting New Strain Relationships

using DrWatson
@quickactivate projectdir()

using SpectralInference
using NewickTree
using HypothesisTests
using MultipleTesting: adjust, Bonferroni, BenjaminiHochberg
using CSV, DataFrames
using StatsBase
using Muon
using Random: shuffle, seed!
using CategoricalArrays: cut, recode!
using LaTeXStrings
using Images

using StatsPlots
theme(:default, grid=false, label=false, tickdir=:out)
include(srcdir("helpers.jl"))

bbdir = datadir("exp_pro", "BB669") |> mkpath;
pdir = plotsdir("strain_variation_tree_and_tests") |> mkpath;
rdir = projectdir("_research", "strain_variation_wilcoxon_tests") |> mkpath

speciescolordf = CSV.read(datadir("exp_raw", "BB669", "subsettreecolors.csv"), DataFrame)
species_color_dict = Dict(k => v for (k, v) in zip(speciescolordf.species_name, speciescolordf.color));

Fig. S11 - Statistical tests through the spectral tree

Given that we are able to create a spectral tree that respects taxonomic relationships and even captures sub-species groupings that corresponds with host environment, we were interested in understanding what genomic signatures were describing these sub-species groupings. However, because these sub-species groups a specifically defined by these genomic features, assumptions of independence typical to statistical tests are violated. Therefore we use a more stringent test to determine whether a given OGG is differentially abundant by using how likely it is for any of the features we are looking at to be significant before deciding if each individual OGG feature is significant.

Specifically to determine whether a given OGG is differentially abundant in a statistically significant way across two groups (‘group A’ and ‘group B’) that arise from a common node in the Spectral Tree, we first generate a ‘reference p-value’ and a ‘null distribution’. The reference p-value signifies the difference in \(OGG_i\) across the two groups (Wilcoxon test). To generate a null distribution of p-values, we create 100 permutations of the distributions between groups A and B that randomly shuffles entries in the two groups. For all OGGs within the Spectral Tree, we compute a distribution of the most significant p-values across all permutations. The Q-value for a given OGG is computed as the number of p-values in the null distribution that are less than the reference p-value (\(P\)) relative to the total number of p-values in the null distribution (equation).

\[Q = \frac{N_{p_{perm}} < N_{p_{ref}}}{N_{p_{perm}}}\]

Filtering to species that have > 20 strain replicates

uniprot = readh5ad(datadir("exp_raw", "UP7047", "2020_02_UP7047.h5ad"))
biobank = readh5mu(datadir("exp_raw", "BB669", "BB669.h5mu"))

┌ Warning: Cannot join columns with the same name because var_names are intersecting.
└ @ Muon /Users/bend/.julia/packages/Muon/UKjAF/src/mudata.jl:367

MuData object 669 ✕ 21475
└ metabolites_foldchange
  AnnData object 669 ✕ 50
└ oggs
  AnnData object 669 ✕ 11248
└ UPorder_oggs
  AnnData object 669 ✕ 10177

upmtx = uniprot.X[:, :]
upusv = svd(upmtx)
bbmtx = biobank["UPorder_oggs"].X[:,:]
size(bbmtx)

(669, 10177)

keptspecies = sort(filter(x-> last(x) >= 20, countmap(biobank.obs.Species)), byvalue=true, rev=true)
filter!(!=("unclassified"), keptspecies)
# species => number of strains belonging to that species

OrderedCollections.OrderedDict{String, Int64} with 12 entries:
  "Phocaeicola vulgatus"         => 88
  "[Ruminococcus] gnavus"        => 41
  "Bacteroides thetaiotaomicron" => 35
  "Anaerostipes hadrus"          => 31
  "Bacteroides uniformis"        => 27
  "unclassified"                 => 26
  "Blautia luti"                 => 24
  "Bifidobacterium breve"        => 24
  "Coprococcus comes"            => 23
  "Dorea formicigenerans"        => 22
  "Blautia wexlerae"             => 21
  "[Eubacterium] rectale"        => 20

I had previously calculated which strains belonged to these species when I saved the biobank dataset

strvarmtx = bbmtx[biobank.obs.kept_species.==1,:];
@show size(strvarmtx)

strvarobs = biobank.obs[biobank.obs.kept_species.==1, :]
strvar_obsnames = strvarobs.Strain_ID

# infer hierarchical relationships of just the species with
# greater than 20 strain replicates
uhat = projectinLSV(strvarmtx, upusv)
dij = spectraldistances(uhat, upusv.S; alpha=1.5, q=.75) ./ size(uniprot, 2)
strvartree_hc = UPGMA_tree(dij)
strvartree = readnw(SpectralInference.newickstring(strvartree_hc, strvar_obsnames))
# ladderize!(strvartree)

## write out tree
open(joinpath(bbdir, "strvar-spitree.nw"), "w") do io
    writenw(io, strvartree)
end

size(strvarmtx) = (356, 10177)

12058

# plot cladogram
strvartree_hc = UPGMA_tree(dij)
subsettreestring = SpectralInference.newickstring(strvartree_hc, strvarobs.Strain_ID)
subsettree = readnw(subsettreestring);
plot(strvartree_hc, 
    size=(800, 900),
    lw=.5,
    yflip=true, 
    xmirror=true,
    xticks=:none,
    permute=(:y, :x),
    grid=false,
    tickdirection=:none,
    rightmargin=5.5Plots.cm, 
    label="",
    framestyle=:grid,
)

# plot annotation rectangles
rectangle(w, h, x, y) = Shape(x .+ [0,w,w,0], y .+ [0,0,h,h])
speciesvector = strvarobs.Species[strvartree_hc.order]
breaks = findall(speciesvector[begin:(end-1)] .!= speciesvector[2:end])[Not([10, 11, 12, 13, 14,15])]
edges = [(s, e) for (s,e) in zip(vcat([0],breaks), vcat(breaks, [length(speciesvector)]))];
rects = [rectangle(2,(e-s),0,s+.5) for (s,e) in zip(vcat([0],breaks), vcat(breaks, [length(speciesvector)]))];
rectspeciescolors = permutedims(speciescolordf.color[indexin(speciesvector[first.(edges).+1], speciescolordf.species_name)]);
fancy_treeplot = plot!(permutedims(rects), fill=0.35, lw=0, c=rectspeciescolors, label="")

# species 
for (i, (k,v)) in enumerate(zip(speciescolordf.species_name, speciescolordf.color))
    yval = median(findall(==(k), speciesvector))
    annotate!(-0.1, yval, text(k, 9, :left, v))
end
annotate!(-0.1, 356, text("Blautia luti (2)", 9, :left, species_color_dict["Blautia luti"]))

fancy_treeplot

Statistically test clades at each node in tree

# functions to compare differences
function cliffs_d(a, b) 
    ans = 0.
    for i in a, j in b
        ans += sign(i - j)
    end
    ans / (length(a) * length(b))
end

function log2FC(a, b) 
    mean(log2.(a .+ 1)) - mean(log2.(b .+ 1))
end

log2FC (generic function with 1 method)

# significance tests across full strain variation tree
seed!(123456789)
tree = strvartree

maxtreeheight = last(maximum(getheights(tree)))
leaves = getleaves(tree)
leafnames = name.(leaves);
treeorder = indexin(leafnames, strvar_obsnames);

ogg_mtx = biobank["oggs"].X[:,:][biobank.obs.kept_species.==1,:]
ogg_mask = vec(mapslices(c->std(c) > 0, ogg_mtx, dims=1))
ogg_mtx = ogg_mtx[treeorder, ogg_mask]
@show size(ogg_mtx)
ogg_names = biobank["oggs"].var_names[ogg_mask]

ogg_freqs = vec(mean(ogg_mtx .> 0, dims=1))

taxon_info = strvarobs[treeorder, :]
taxon_info.species_donor = join.(eachrow(strvarobs[treeorder, [:Species, :Donor]]), " ")


isolate_ids = getleafnames(tree)
testresults = DataFrame()


for node in prewalk(tree)
    #  if group size would be less than 1 skip node
    (isleaf(node) || any(isleaf.(node.children))) && continue
    
    # find comparison groups
    grp1_idx = indexin(getleafnames(node.children[1]), isolate_ids)
    grp2_idx = indexin(getleafnames(node.children[2]), isolate_ids)
    
    pvals = mapslices(ogg_mtx, dims=1) do column
        pvalue(MannWhitneyUTest(column[grp1_idx], column[grp2_idx]))
    end |> vec
    
    # make null
    seed!(1234)
    null_qs = map(1:100) do i
        n = length(grp1_idx)
        m = length(grp2_idx)
        perm_idx = shuffle(vcat(grp1_idx, grp2_idx))
        qs = mapslices(ogg_mtx, dims=1) do column_perm
            q = pvalue(MannWhitneyUTest(column_perm[perm_idx[1:n]], column_perm[perm_idx[n+1:end]]))
        end
        minimum(qs)
    end

    qvals = mapslices(ogg_mtx, dims=1) do column
        pval = pvalue(MannWhitneyUTest(column[grp1_idx], column[grp2_idx]))
        mean(null_qs .< pval)
    end |> vec
    
    effects = mapslices(ogg_mtx, dims=1) do column
        abs(mean(column[grp1_idx]) - mean(column[grp2_idx]) / std(column))
    end |> vec

    log_effects = mapslices(ogg_mtx, dims=1) do column
        logcol = log2.(column.+1)
        abs(mean(logcol[grp1_idx]) - mean(logcol[grp2_idx]) / std(logcol))
    end |> vec
    
    log2FCs = mapslices(ogg_mtx, dims=1) do column
        log2FC(column[grp1_idx], column[grp2_idx])
    end |> vec

    cliffs_ds = mapslices(ogg_mtx, dims=1) do column
        cliffs_d(column[grp1_idx], column[grp2_idx])
    end |> vec

    grp1_prp_exp = mapslices(ogg_mtx, dims=1) do column
        mean(column[grp1_idx] .> 0)
    end |> vec

    grp2_prp_exp = mapslices(ogg_mtx, dims=1) do column
        mean(column[grp2_idx] .> 0)
    end |> vec
    
    testresults = vcat(testresults, 
        DataFrame(
            :nodeids => id(node),
            :nodeheight =>  NewickTree.height(node),
            :nodedepth =>  maxtreeheight - NewickTree.height(node),
            :grp1_N => length(grp1_idx),
            :grp2_N => length(grp2_idx),
            :grp1_phylum_mode => mode(taxon_info.Phylum[grp1_idx]),
            :grp2_phylum_mode => mode(taxon_info.Phylum[grp2_idx]),
            :grp1_species_mode => mode(taxon_info.Species[grp1_idx]),
            :grp2_species_mode => mode(taxon_info.Species[grp2_idx]),
            :grp1_species_donor_mode => mode(taxon_info.species_donor[grp1_idx]),
            :grp2_species_donor_mode => mode(taxon_info.species_donor[grp2_idx]),
            :ogg_name => ogg_names,
            :ogg_idx => collect(axes(ogg_mtx, 2)),
            :ogg_freqs => ogg_freqs,
            :grp1_prp_exp => grp1_prp_exp,
            :grp2_prp_exp => grp2_prp_exp,
            :effectsize => effects,
            :logeffectsize => log_effects,
            :log2FC => log2FCs,
            :cliffs_d => cliffs_ds,
            :pvals => pvals,
            :qvals => qvals,
        )
    )
end
testresults[!, "pval_BH"] .= adjust(testresults.pvals, BenjaminiHochberg());
testresults[!, "pval_Bon"] .= adjust(testresults.pvals, Bonferroni());
testresults[!, "qval_BH"] .= adjust(testresults.qvals, BenjaminiHochberg());

tree_discretization = cut(testresults.nodedepth, [0, 2, 4, 5])
recode!(tree_discretization, 
    "[4, 5)"=>"phylum level", 
    "[2, 4)"=>"species level",
    "[0, 2)"=>"strain level")
testresults[!, "tree_level"] = tree_discretization
testresults;

size(ogg_mtx) = (356, 5449)

countmap(testresults.tree_level)

Dict{CategoricalArrays.CategoricalValue{String, UInt32}, Int64} with 3 entries:
  "species level" => 59939
  "strain level"  => 457716
  "phylum level"  => 5449

CSV.write(joinpath(rdir, "full_ogg_wilcoxon_testresults_on_big10species.csv"), testresults)
testresults |> 
    df->filter(:pval_BH => <(0.05), df) |>
    df->CSV.write(joinpath(rdir, "significant_BH_ogg_wilcoxon_testresults_on_big10species.csv"), df)
testresults |>
    df->filter(:pval_Bon => <(0.05), df) |>
    df->CSV.write(joinpath(rdir, "significant_Bon_ogg_wilcoxon_testresults_on_big10species.csv"), df)
testresults |>
    df->filter(:qval_BH => <(0.05), df) |>
    df->CSV.write(joinpath(rdir, "significant_Qbh_ogg_wilcoxon_testresults_on_big10species.csv"), df)

"/Users/bend/projects/Doran_etal_2023/_research/strain_variation_wilcoxon_tests/significant_Qbh_ogg_wilcoxon_testresults_on_big10species.csv"

Filtering to most significant results

testresults = CSV.read(joinpath(rdir, "full_ogg_wilcoxon_testresults_on_big10species.csv"), DataFrame);

sigtestresults = filter(:qval_BH => <(0.05), testresults) |>
    df -> filter(:logeffectsize => >(1), df) |>
    df -> filter(:effectsize => >(1), df) |>
    df -> filter(:log2FC => x->abs.(x) .> 1 , df);

@show sigtestresults.ogg_name |> unique |> length
@show sigtestresults.nodeids |> unique |> length
@show sum(sigtestresults.tree_level .== "phylum level")
@show sum(sigtestresults.tree_level .== "species level")
@show sum(sigtestresults.tree_level .== "strain level");

(sigtestresults.ogg_name |> unique) |> length = 1530
(sigtestresults.nodeids |> unique) |> length = 66
sum(sigtestresults.tree_level .== "phylum level") = 415
sum(sigtestresults.tree_level .== "species level") = 1117
sum(sigtestresults.tree_level .== "strain level") = 1070

# how many OGGs are still duplicated?
sigtestresults |> 
    df -> groupby(df, :ogg_name) |>
    df -> subset(df, :log2FC => x -> abs.(x) .== maximum(abs.(x))) |>
    df -> countmap(df.ogg_name) |>
    cm -> sum(values(cm) .> 1)

27

# for these remaining OGGs select the shallowest significant example
uniqueOGGresults = sigtestresults |> 
    df -> groupby(df, :ogg_name) |>
    df -> subset(df, :log2FC => x -> abs.(x) .== maximum(abs.(x))) |>
    df -> groupby(df, :ogg_name) |>
    df -> subset(df, :nodeheight => x -> x .== minimum(x));
@show nrow(uniqueOGGresults) == length(unique(uniqueOGGresults.ogg_name))
@show sum(uniqueOGGresults.tree_level .== "phylum level")
@show sum(uniqueOGGresults.tree_level .== "species level")
@show sum(uniqueOGGresults.tree_level .== "strain level");

nrow(uniqueOGGresults) == length(unique(uniqueOGGresults.ogg_name)) = true
sum(uniqueOGGresults.tree_level .== "phylum level") = 273
sum(uniqueOGGresults.tree_level .== "species level") = 676
sum(uniqueOGGresults.tree_level .== "strain level") = 581

CSV.write(joinpath(rdir, "unique_sig_ogg_wilcoxon_testresults_on_big10species.csv"), uniqueOGGresults)

"/Users/bend/projects/Doran_etal_2023/_research/strain_variation_wilcoxon_tests/unique_sig_ogg_wilcoxon_testresults_on_big10species.csv"

Phylum level significant orthologs

ogg_plotting_order = uniqueOGGresults |>
    df->subset(df,  :tree_level => x-> x.==("phylum level"), :grp1_N => x-> x.>=3, :grp2_N => x-> x.>=3) |>
    df->subset(df, :nodeids => (x->get.(Ref(countmap(x)), x, 0) .> 10)) |>
    df->sort(df, [:nodeids, :log2FC]);

hmapplot = heatmap(log2.(ogg_mtx[:, ogg_plotting_order.ogg_idx] .+ 1), 
# hmapplot = heatmap(log2.(ogg_mtx[:, uniqueOGGresults.ogg_idx]),
    title="Phylum level tests; " * L"Q_{BH} < 0.05, |\log_2 FC| > 1",
    yticks=false,
    # xticks=(1:nrow(uniqueOGGresults), uniqueOGGresults.ogg_name),
    xticks=[1, findall(diff(ogg_plotting_order.nodeids) .!= 0)..., nrow(ogg_plotting_order)],
    xrotation=45,
    ylims=(.5,356.5),
    framestyle=:grid,
    cticks=(1:4, string.(2 .^ 1:4)),
    c=cgrad(:bilbao, rev=true),
    leftmargin=0Plots.mm,
);

rectangle(w, h, x, y) = Shape(x .+ [0,w,w,0], y .+ [0,0,h,h])
leafnames = getleafnames(strvartree)
oggrects = map(unique(ogg_plotting_order.nodeids)) do nid
    node = filter(x-> id(x)==(nid), prewalk(strvartree))[1]
    idx_taxa = indexin(getleafnames(node), leafnames)
    noderesults = filter(:nodeids => ==(nid), ogg_plotting_order)
    idx_oggs = indexin(noderesults.ogg_name, ogg_plotting_order.ogg_name)
    x = minimum(idx_oggs)
    y = minimum(idx_taxa)
    width = maximum(idx_oggs) - minimum(idx_oggs)
    height = maximum(idx_taxa) - minimum(idx_taxa)
    rectangle(width, height, x, y)
end

plot!(permutedims(oggrects), fill=0.0, lw=1.5, linecolor="red", label="")

layout = @layout [a{.2w} b]
plot(plot(fancy_treeplot, title="SPI tree"), hmapplot; layout, xrotation=45, size=(1000,600), link=:y, legend=:none, margin=3Plots.Measures.mm)

savefig(joinpath(pdir, "significantOGGS_pernode_phylumlevel_treeandheatmap_coloredboxes.pdf"))

"/Users/bend/projects/Doran_etal_2023/plots/strain_variation_tree_and_tests/significantOGGS_pernode_phylumlevel_treeandheatmap_coloredboxes.pdf"

Species level significant orthologs

ogg_plotting_order = uniqueOGGresults |>
    df->subset(df,  :tree_level => x-> x.==("species level"), :grp1_N => x-> x.>=3, :grp2_N => x-> x.>=3) |>
    df->subset(df, :nodeids => (x->get.(Ref(countmap(x)), x, 0) .> 10)) |>
    df->sort(df, [:nodeids, :log2FC]);

hmapplot = heatmap(log2.(ogg_mtx[:, ogg_plotting_order.ogg_idx] .+ 1), 
# hmapplot = heatmap(log2.(ogg_mtx[:, uniqueOGGresults.ogg_idx]),
    title="Species level tests; " * L"Q_{BH} < 0.05, |\log_2 FC| > 1",
    yticks=false,
    # xticks=(1:nrow(uniqueOGGresults), uniqueOGGresults.ogg_name),
    xticks=[1, findall(diff(ogg_plotting_order.nodeids) .!= 0)..., nrow(ogg_plotting_order)],
    xrotation=45,
    ylims=(.5,356.5),
    framestyle=:grid,
    cticks=(1:4, string.(2 .^ 1:4)),
    c=cgrad(:bilbao, rev=true),
    leftmargin=0Plots.mm,
);

rectangle(w, h, x, y) = Shape(x .+ [0,w,w,0], y .+ [0,0,h,h])
leafnames = getleafnames(strvartree)
oggrects = map(unique(ogg_plotting_order.nodeids)) do nid
    node = filter(x-> id(x)==(nid), prewalk(strvartree))[1]
    idx_taxa = indexin(getleafnames(node), leafnames)
    noderesults = filter(:nodeids => ==(nid), ogg_plotting_order)
    idx_oggs = indexin(noderesults.ogg_name, ogg_plotting_order.ogg_name)
    x = minimum(idx_oggs)
    y = minimum(idx_taxa)
    width = maximum(idx_oggs) - minimum(idx_oggs)
    height = maximum(idx_taxa) - minimum(idx_taxa)
    rectangle(width, height, x, y)
end

plot!(permutedims(oggrects), fill=0.0, lw=1.5, linecolor="aqua", label="")

layout = @layout [a{.2w} b]
plot(plot(fancy_treeplot, title="SPI tree"), hmapplot; layout, xrotation=45, size=(1000,600), link=:y, legend=:none, margin=3Plots.Measures.mm)

savefig(joinpath(pdir, "significantOGGS_pernode_specieslevel_treeandheatmap_coloredboxes.pdf"))

"/Users/bend/projects/Doran_etal_2023/plots/strain_variation_tree_and_tests/significantOGGS_pernode_specieslevel_treeandheatmap_coloredboxes.pdf"

Strain level significant orthologs

ogg_plotting_order = uniqueOGGresults |>
    df->subset(df,  :tree_level => x-> x.==("strain level"), :grp1_N => x-> x.>=3, :grp2_N => x-> x.>=3) |>
    df->subset(df, :nodeids => (x->get.(Ref(countmap(x)), x, 0) .> 10)) |>
    df->sort(df, [:nodeids, :log2FC]);

hmapplot = heatmap(log2.(ogg_mtx[:, ogg_plotting_order.ogg_idx] .+ 1), 
# hmapplot = heatmap(log2.(ogg_mtx[:, uniqueOGGresults.ogg_idx]),
    title="Strain level tests; " * L"Q_{BH} < 0.05, |\log_2 FC| > 1",
    yticks=false,
    # xticks=(1:nrow(uniqueOGGresults), uniqueOGGresults.ogg_name),
    xticks=[1, findall(diff(ogg_plotting_order.nodeids) .!= 0)..., nrow(ogg_plotting_order)],
    xrotation=45,
    ylims=(.5,356.5),
    framestyle=:grid,
    cticks=(1:4, string.(2 .^ 1:4)),
    c=cgrad(:bilbao, rev=true),
    leftmargin=0Plots.mm,
);

rectangle(w, h, x, y) = Shape(x .+ [0,w,w,0], y .+ [0,0,h,h])
leafnames = getleafnames(strvartree)
oggrects = map(unique(ogg_plotting_order.nodeids)) do nid
    node = filter(x-> id(x)==(nid), prewalk(strvartree))[1]
    idx_taxa = indexin(getleafnames(node), leafnames)
    noderesults = filter(:nodeids => ==(nid), ogg_plotting_order)
    idx_oggs = indexin(noderesults.ogg_name, ogg_plotting_order.ogg_name)
    x = minimum(idx_oggs)
    y = minimum(idx_taxa)
    width = maximum(idx_oggs) - minimum(idx_oggs)
    height = maximum(idx_taxa) - minimum(idx_taxa)
    rectangle(width, height, x, y)
end

plot!(permutedims(oggrects), fill=0.0, lw=1.5, linecolor="lightgreen", label="")
# vline!(findall(diff(ogg_plotting_order.nodeids) .> 0) .- .5, lw=1, c=:black, label="")

layout = @layout [a{.2w} b]
plot(plot(fancy_treeplot, title="SPI tree"), hmapplot; layout, xrotation=45, size=(1000,600), link=:y, legend=:none, margin=3Plots.Measures.mm)

savefig(joinpath(pdir, "significantOGGS_pernode_strainlevel_treeandheatmap_coloredboxes.pdf"))

"/Users/bend/projects/Doran_etal_2023/plots/strain_variation_tree_and_tests/significantOGGS_pernode_strainlevel_treeandheatmap_coloredboxes.pdf"

zoom-ins of strain level significant orthologs

### Strain level zoomin P. vulgatus
ogg_plotting_order.nodeids |> unique
ogg_plotting_order = uniqueOGGresults |>
    df->subset(df,  :tree_level => x-> x.==("strain level"), :grp1_N => x-> x.>=3, :grp2_N => x-> x.>=3) |>
    df->subset(df, :nodeids => (x->get.(Ref(countmap(x)), x, 0) .> 10)) |>
    df->sort(df, [:nodeids, :log2FC]) |>
    df->filter(:nodeids => ==(136), df)
nid = 136
node = filter(x-> id(x)==(nid), prewalk(strvartree))[1]
idx_taxa = indexin(getleafnames(node), leafnames)

hmapplot = heatmap(log2.(ogg_mtx[idx_taxa, ogg_plotting_order.ogg_idx] .+ 1), 
# hmapplot = heatmap(log2.(ogg_mtx[:, uniqueOGGresults.ogg_idx]),
    title="P. vulgatus (zoomin); " * L"Q_{BH} < 0.05, |\log_2 FC| > 1",
    yticks=false,
    xticks=(1:nrow(ogg_plotting_order), ogg_plotting_order.ogg_name),
    # xtickfontsize=1,
    xrotation=90,
    # ylims=(.5,356.5),
    tickdirection=:out,
    framestyle=:grid,
    cticks=(1:4, string.(2 .^ 1:4)),
    clims=(0, 6.1),
    c=cgrad(:bilbao, rev=true),
    size=(550,550),
    leftmargin=0Plots.mm,
    grid=false
)

### Strain level zoomin C. comes
ogg_plotting_order.nodeids |> unique
ogg_plotting_order = uniqueOGGresults |>
    df->subset(df,  :tree_level => x-> x.==("strain level"), :grp1_N => x-> x.>=3, :grp2_N => x-> x.>=3) |>
    df->subset(df, :nodeids => (x->get.(Ref(countmap(x)), x, 0) .> 10)) |>
    df->sort(df, [:nodeids, :log2FC]) |>
    df->filter(:nodeids => ==(523), df)
nid = 523
node = filter(x-> id(x)==(nid), prewalk(strvartree))[1]
idx_taxa = indexin(getleafnames(node), leafnames)

hmapplot = heatmap(log2.(ogg_mtx[idx_taxa, ogg_plotting_order.ogg_idx] .+ 1), 
    title="C. comes (zoomin); " * L"Q_{BH} < 0.05, |\log_2 FC| > 1",
    yticks=false,
    xticks=(1:nrow(ogg_plotting_order), ogg_plotting_order.ogg_name),
    xrotation=90,
    tickdirection=:out,
    framestyle=:grid,
    cticks=(1:4, string.(2 .^ 1:4)),
    clims=(0, 6.1),
    c=cgrad(:bilbao, rev=true),
    size=(550,550),
    leftmargin=0Plots.mm,
    grid=false
)

Main - Explore differences within E. rectale

targetid = "MSK.16.22"
strvar_leaves = getleaves(strvartree)
basenode = strvar_leaves[findfirst(n->name(n) == targetid, strvar_leaves)]
subtree = readnw(nwstr(nthparent(basenode, 3)))
rectale_idxs = indexin(getleafnames(subtree),strvarobs.Strain_ID)


tiplabels = strvarobs.Donor[rectale_idxs]
idmapping = Dict(k=>v for (k,v) in zip(getleafnames(subtree), tiplabels))
rename_treeleaves!(subtree, idmapping)

rectale_donor_dict = Dict(
    "MSK.22"=>"#2B3990", 
    "MSK.17"=>"#6682C1", 
    "MSK.16"=>"#FBB040", 
    "MSK.13"=>"#c9a964", 
    "MSK.9"=>"#A97C50",
)

rectale_linecolors = map(prewalk(subtree)) do node; rectale_donor_dict
    if length(unique(getleafnames(node))) == 1
        get(rectale_donor_dict, mode(getleafnames(node)), :grey)
    else
        :grey
    end
end |> x->x[2:end] |> permutedims

plot(subtree,
    c=rectale_linecolors,
    # alpha=.5,
    lw=3, fs=2,
    size=(400,600),
    rightmargin=2Plots.cm,
    leftmargin=10Plots.mm,
    ylabel="E. rectale",
    framestyle=:grid,
    ticks=false,
)
# donor annotations 
for (i, (k,v)) in enumerate(rectale_donor_dict)
    yvals = findall(==(k), getleafnames(subtree))
    yval = median(yvals)
    annotate!(1.8, yval, text("$k\n(n=$(length(yvals)))", 9, :center, v))
end
p1 = plot!()

ogg_plotting_order = uniqueOGGresults |>
    # select test at root of rectale sub tree
    df->subset(df, :nodeids => x-> x.==(id(nthparent(basenode, 3)))) |>
    df->select(df, [:nodeids, :log2FC, :grp1_N, :grp2_N, :ogg_name, :ogg_idx]) |>
    # add in descriptions of each ogg
    df->leftjoin(df, uniprot.var[:, [:og, :description]], on = :ogg_name => :og) |>
    # sort oggs based on abundence per group
    df->sort(df, [:nodeids, :log2FC]) |>
    # get shorter descriptions for lables
    df->coalesce.(df, "") |>
    df->transform(df, 
        :description => ByRow(x->first(replace.(x, r"\(.*\)"=>""), 75)) => :ogg_labels
    );

# more abundant in MSK.17 and MSK.22
subset(ogg_plotting_order,
    :log2FC => ByRow(<(0)),
    :description => ByRow(!=(""))
)

17×8 DataFrame

Row	nodeids	log2FC	grp1_N	grp2_N	ogg_name	ogg_idx	description	ogg_labels
	Int64	Float64	Int64	Int64	String15	Int64	String	String
1	433	-2.52345	7	13	COG1345	3757	Required for morphogenesis and for the elongation of the flagellar filament by facilitating polymerization of the flagellin monomers at the tip of growing filament. Forms a capping structure, which prevents flagellin subunits (transported through the central channel of the flagellum) from leaking out without polymerization at the distal end	Required for morphogenesis and for the elongation of the flagellar filament
2	433	-2.26897	7	13	COG1344	3756	bacterial-type flagellum-dependent cell motility	bacterial-type flagellum-dependent cell motility
3	433	-2.03848	7	13	COG5444	5401	nuclease activity	nuclease activity
4	433	-2.03047	7	13	COG5001	5338	cyclic-guanylate-specific phosphodiesterase activity	cyclic-guanylate-specific phosphodiesterase activity
5	433	-2.0	7	13	COG3210	4625	domain, Protein	domain, Protein
6	433	-1.94674	7	13	COG3547	4769	Transposase (IS116 IS110 IS902 family)	Transposase
7	433	-1.8723	7	13	COG4495	5124	Domain of unknown function (DUF4176)	Domain of unknown function
8	433	-1.74526	7	13	COG1442	3821	lipopolysaccharide 3-alpha-galactosyltransferase activity	lipopolysaccharide 3-alpha-galactosyltransferase activity
9	433	-1.58496	7	13	COG4786	5238	Flagellar basal body rod	Flagellar basal body rod
10	433	-1.50843	7	13	COG5279	5367	protein involved in cytokinesis, contains TGc (transglutaminase protease-like) domain	protein involved in cytokinesis, contains TGc domain
11	433	-1.38462	7	13	COG1776	4031	chemotaxis	chemotaxis
12	433	-1.35998	7	13	COG2944	4514	sequence-specific DNA binding	sequence-specific DNA binding
13	433	-1.29925	7	13	COG0842	3475	Transport permease protein	Transport permease protein
14	433	-1.29925	7	13	COG1516	3869	flagellar protein fliS	flagellar protein fliS
15	433	-1.22499	7	13	COG0835	3469	chemotaxis	chemotaxis
16	433	-1.22499	7	13	COG2186	4263	Transcriptional regulator	Transcriptional regulator
17	433	-1.22499	7	13	COG2198	4271	Histidine kinase	Histidine kinase

# more abundant in MSK.9, MSK.13, and MSK.16
subset(ogg_plotting_order,
    :log2FC => ByRow(>(0)),
    :description => ByRow(!=(""))
)

5×8 DataFrame

Row	nodeids	log2FC	grp1_N	grp2_N	ogg_name	ogg_idx	description	ogg_labels
	Int64	Float64	Int64	Int64	String15	Int64	String	String
1	433	1.16713	7	13	COG5015	5347	Pyridoxamine 5'-phosphate oxidase	Pyridoxamine 5'-phosphate oxidase
2	433	1.32193	7	13	COG3534	4763	alpha-L-arabinofuranosidase	alpha-L-arabinofuranosidase
3	433	1.64425	7	13	COG3378	4691	Phage plasmid primase P4 family	Phage plasmid primase P4 family
4	433	1.92997	7	13	COG0286	2997	site-specific DNA-methyltransferase (adenine-specific) activity	site-specific DNA-methyltransferase activity
5	433	2.99151	7	13	COG0732	3379	type I restriction modification DNA specificity domain	type I restriction modification DNA specificity domain

pltmtx = ogg_mtx[indexin(strvarobs.Strain_ID, getleafnames(strvartree)), ogg_plotting_order.ogg_idx][rectale_idxs, :]
p2 = heatmap(log2.(pltmtx .+ 1), 
    yticks=false,
    xticks=(1:nrow(ogg_plotting_order), ogg_plotting_order.ogg_labels),
    tickdirection=:out,
    cticks=(1:4, string.(2 .^ 1:4)),
    c=cgrad(:bilbao, rev=true),
    leftmargin=0Plots.mm,
    rightmargin=4Plots.cm,
    topmargin=1.5Plots.cm,
    grid=false,
    framestyle=:box,
    xmirror=true,
    xrotation=45,
);

plot(p1, p2, layout=@layout([a{.25w} b]), link=:y, size=(1000, 800))

Validation experiment

We noted that groups of strains from donors 22 and 17 had many orthologs related to the flagellum, that were absent from donors 9, 13, and 16. Conversely donors 9, 13, and 16 had orthologs related to viral (bacteriophage) infection. It suggested that in donors 9, 13, and 16 there had been an infection of populations that caused these strains to lose their flagellum and thus motility, likely to become more resistant to this bacteriophage.

We just wanted to test whether these strains where indeed non-motile, to verify that we were detecting accurate and functional differences between these strains. So we grew a sample of these strain in BHIS media, took pictures at the 24 hour mark and a time-series of optical density measurements (OD600).

earlydf = CSV.read(datadir("exp_raw", "OD600_motility", "earlygrowth_OD600.tsv"), DataFrame, delim="\t") |>
    df -> stack(df, 2:5) |>
    df -> rename(df, :variable => :time, :value => :OD600);
earlydf = earlydf |>
    df -> transform(df,
        :ID => (x->join.(first.(split.(x, "."), 2), ".")) => :msk_id,
        :ID => (x->last.(split.(x, "."))) => :replicate,
    );

vortexdf = CSV.read(datadir("exp_raw", "OD600_motility", "postvortex_OD600.tsv"), DataFrame, delim="\t") |>
    df -> stack(df, 2:8) |>
    df -> rename(df, :variable => :time, :value => :OD600);
vortexdf = vortexdf |>
    df -> transform(df,
        :ID => (x->join.(first.(split.(x, "."), 2), ".")) => :msk_id,
        :ID => (x->last.(split.(x, "."))) => :replicate,
    );
vortexpltdf = vortexdf |>
    df->groupby(df, [:time, :msk_id]) |>
    df->combine(df, 
        :OD600 => mean,
        :OD600 => std,
    );

fulldf = vcat(earlydf, vortexdf)
# fulldf = filter(:msk_id => x->startswith(x, r"(9|17|22|neg)"), fulldf)
fullpltdf = fulldf |>
    df->groupby(df, [:time, :msk_id]) |>
    df->combine(df, 
        :OD600 => mean,
        :OD600 => std,
    );
fullpltdf.OD600_std .= replace(fullpltdf.OD600_std, NaN => 0.); # set correct std for single negative control

cmap=Dict(
    "neg" => :grey,
    "9.13" => "#A97C50",
    "9.15" => "#A97C50",
    "16.22" => "#FBB040",
    "17.70" => "#6682C1",
    "17.78" => "#6682C1",
    "22.92" => "#2B3990"
)

plot(
    ylabel="OD600",
    legend=:outerright,
    xrotation=45,
    margin=5Plots.Measures.mm,
    size=(700,300),
    bottommargins=10Plots.mm,
)
@df fulldf scatter!(
    :time, :OD600,
    group=:ID, 
    c=[cmap[id] for id in :msk_id],
    label="", markersize=3, markerstrokewidth=0,
)
vline!([4], c=:black, lw=.5, linestyle=:dash)
@df fullpltdf plot!(
    :time, :OD600_mean,
    group=:msk_id, ribbon=:OD600_std,
    c=[cmap[id] for id in :msk_id],
    fillalpha=.25,
)
annotate!(4.8, .27, text("vortex"))

savefig(joinpath(pdir, "motility_experiment_plot.pdf"))

"/Users/bend/projects/Doran_etal_2023/plots/strain_variation_tree_and_tests/motility_experiment_plot.pdf"

img = load(datadir("exp_raw", "OD600_motility", "swimmers_after24h.jpg"))
ysize = floor.(Int, quantile(1:size(img,1), [0.2, 0.77])) |> x->x[1]:x[2]
xsize = floor.(Int, quantile(1:size(img,2), [0.15, 0.73])) |> x->x[1]:x[2]
img[ysize, xsize]

img = load(datadir("exp_raw", "OD600_motility", "nonswimmers_after24h.jpg"))
ysize = floor.(Int, quantile(1:size(img,1), [0.2, 0.77])) |> x->x[1]:x[2]
xsize = floor.(Int, quantile(1:size(img,2), [0.2, 0.71])) |> x->x[1]:x[2]
img[ysize, xsize]

How conserved are these features?

The flagellum is generally highly conserved, it is remarkable that phage seemed to be involved with deleting such seemingly important machinery from the organism.

Thus, we examined the conservation pattern of the 12 annotated gene groups that were absent in E. rectale strains isolated from donors 9, 13, 16 but present in strains isolated from donors 17 and 22 across the entire Spectral Tree

uniprot = readh5ad(datadir("exp_raw", "UP7047", "2020_02_UP7047.h5ad"))
upmtx = uniprot.X[:,:];
uptree = readnw(readline(projectdir("_research", "UP7047_neighborjoined_spitree", "2020_02_UP7047-supporttree_50pct.nw")));
leafnodes = getleaves(uptree);
treeorder = indexin(getleafnames(uptree), uniprot.obs_names);

species_name = "rectale"
foundtaxa = uniprot.obs[contains.(uniprot.obs.Species, species_name), [:Kingdom, :Phylum, :Class, :Order, :Family, :Genus, :Species]]
@info "string: '$(species_name)' finds $(nrow(foundtaxa)) taxa:\n$(foundtaxa)"

┌ Info: string: 'rectale' finds 1 taxa:
│ 1×7 DataFrame
│  Row │ Kingdom   Phylum      Class       Order          Family           Genus   Species
│      │ String    String      String      String         String           String  String
│ ─────┼─────────────────────────────────────────────────────────────────────────────────────────────────
│    1 │ Bacteria  Firmicutes  Clostridia  Clostridiales  Lachnospiraceae          [Eubacterium] rectale
└ @ Main /Users/bend/projects/Doran_etal_2023/notebooks/jl_notebook_cell_df34fa98e69747e1a8f8a730347b8e2f_Y101sZmlsZQ==.jl:3

test_oggs = subset(ogg_plotting_order,
    :description => ByRow(!=("")), # drop unannotated oggs
    :description => ByRow(!=("domain, Protein")), # drop unannotated oggs
    :log2FC => ByRow(<(0)), # drop oggs increased in the presence of phage
).ogg_name
uniprot.var[indexin(sort(test_oggs), uniprot.var_names), [:og, :description]]

16×2 DataFrame

Row	og	description
	String	String
1	COG0835	chemotaxis
2	COG0842	Transport permease protein
3	COG1344	bacterial-type flagellum-dependent cell motility
4	COG1345	Required for morphogenesis and for the elongation of the flagellar filament by facilitating polymerization of the flagellin monomers at the tip of growing filament. Forms a capping structure, which prevents flagellin subunits (transported through the central channel of the flagellum) from leaking out without polymerization at the distal end
5	COG1442	lipopolysaccharide 3-alpha-galactosyltransferase activity
6	COG1516	flagellar protein fliS
7	COG1776	chemotaxis
8	COG2186	Transcriptional regulator
9	COG2198	Histidine kinase
10	COG2944	sequence-specific DNA binding
11	COG3547	Transposase (IS116 IS110 IS902 family)
12	COG4495	Domain of unknown function (DUF4176)
13	COG4786	Flagellar basal body rod
14	COG5001	cyclic-guanylate-specific phosphodiesterase activity
15	COG5279	protein involved in cytokinesis, contains TGc (transglutaminase protease-like) domain
16	COG5444	nuclease activity

# oggs from E. rectale test that are in Uniprot
chosen_oggs_idx = indexin(test_oggs, uniprot.var_names);
# make presence/absence matix `subset_ogg_mtx`
subset_ogg_mtx = upmtx[:, chosen_oggs_idx] .> 0;
chosen_oggs = uniprot.var.og[chosen_oggs_idx]
@show chosen_oggs;

chosen_oggs = ["COG1345", "COG1344", "COG5444", "COG5001", "COG3547", "COG4495", "COG1442", "COG4786", "COG5279", "COG1776", "COG2944", "COG0842", "COG1516", "COG0835", "COG2186", "COG2198"]

# collect parent nodes back to pseudo-root of uniprot tree
species_node_idx = findall(contains.(uniprot.obs.Species[treeorder], species_name))[1]
speciesnode = leafnodes[species_node_idx]
depthofspecies = network_distance(uptree, speciesnode)
speciesparents = []
node = speciesnode
for i in 1:depthofspecies
    node = parent(node)
    push!(speciesparents, node)
end
cladesizes = length.(getleafnames.(speciesparents))
@info "# of taxa per clade = $(cladesizes)"

┌ Info: # of taxa per clade = [3, 10, 23, 101, 216, 265, 380, 600, 733, 1022, 1065, 2376, 6709, 7047]
└ @ Main /Users/bend/projects/Doran_etal_2023/notebooks/jl_notebook_cell_df34fa98e69747e1a8f8a730347b8e2f_Y104sZmlsZQ==.jl:12

# only 1 E. rectale in uniprot
sum(==("Agathobacter"), uniprot.obs.Genus)

1

Local subtree of E. rectale in Uniprot tree

subset_tree = readnw(nwstr(speciesparents[3]))
for leaf in getleaves(subset_tree)
    NewickTree.setname!(leaf, only(uniprot.obs.Species[uniprot.obs.Proteome_ID .== name(leaf)]) )
end
plot(subset_tree, rightmargin=6Plots.cm, fs=9, size=(500,600))

Larger subtree of E. rectale in Uniprot tree

subset_tree = readnw(nwstr(speciesparents[4]))
for leaf in getleaves(subset_tree)
    NewickTree.setname!(leaf, only(uniprot.obs.Species[uniprot.obs.Proteome_ID .== name(leaf)]) )
end
plot(subset_tree, rightmargin=4Plots.cm, fs=5, size=(400,600))

bblsvs = projectinLSV(bbmtx, upusv);

rectale_mask = occursin.("rectale", biobank.obs.Species);

joined_dij = spectraldistances([upusv.U; bblsvs[rectale_mask, :]], upusv.S, alpha=1.5, q=0.75);

rectale_dij = joined_dij[(size(upusv.U,1)+1):end, 1:(end-sum(rectale_mask))]
rectale_upnns = first.(sortperm.(eachrow(rectale_dij)), 5);

hcat(biobank.obs[rectale_mask, [:Strain_ID, :NCBI_Species]],
    DataFrame(getindex.(Ref(uniprot.obs.Species), rectale_upnns) |> stack |> permutedims, [:nn_1, :nn_2, :nn_3, :nn_4, :nn_5]),
     DataFrame(;
    # uniprot_nearest_neighbor_Species=uniprot.obs.Species[rectale_upnns],
    # uniprot_nearest_neighbor_Phylum=uniprot.obs.Phylum[rectale_upnns],
    )
)

20×7 DataFrame

Row	Strain_ID	NCBI_Species	nn_1	nn_2	nn_3	nn_4	nn_5
	String	String	String	String	String	String	String
1	MSK.9.13	[Eubacterium] rectale	[Eubacterium] rectale	Aeromonas sp. w55	Photobacterium damselae	Cyanobacteria bacterium	Bacillus sp. UFRGS-B20
2	MSK.13.59	[Eubacterium] rectale	Photobacterium damselae	Cyanobacteria bacterium	[Eubacterium] rectale	Candidatus Nasuia deltocephalinicola	Aeromonas sp. w55
3	MSK.16.22	[Eubacterium] rectale	[Eubacterium] rectale	Cyanobacteria bacterium	Candidatus Vidania fulgoroideae	Candidatus Carsonella ruddii	Bacillus sp. UFRGS-B20
4	MSK.9.15	[Eubacterium] rectale	Aeromonas sp. w55	Photobacterium damselae	[Eubacterium] rectale	Cyanobacteria bacterium	Candidatus Tremblaya princeps
5	MSK.13.48	[Eubacterium] rectale	Photobacterium damselae	Cyanobacteria bacterium	[Eubacterium] rectale	Candidatus Vidania fulgoroideae	Candidatus Nasuia deltocephalinicola
6	MSK.13.50	[Eubacterium] rectale	Photobacterium damselae	Cyanobacteria bacterium	[Eubacterium] rectale	Candidatus Nasuia deltocephalinicola	Candidatus Vidania fulgoroideae
7	MSK.17.42	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Photobacterium damselae	Cyanobacteria bacterium	Opitutae bacterium SCGC AG-212-L18
8	MSK.17.3	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Photobacterium damselae	Cyanobacteria bacterium	Opitutae bacterium SCGC AG-212-L18
9	MSK.17.13	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Photobacterium damselae	Cyanobacteria bacterium	Opitutae bacterium SCGC AG-212-L18
10	MSK.17.19	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Cyanobacteria bacterium	Photobacterium damselae	Opitutae bacterium SCGC AG-212-L18
11	MSK.17.57	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Cyanobacteria bacterium	Photobacterium damselae	Opitutae bacterium SCGC AG-212-L18
12	MSK.22.19	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Cyanobacteria bacterium	Photobacterium damselae	Candidatus Tremblaya princeps
13	MSK.17.70	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Photobacterium damselae	Opitutae bacterium SCGC AG-212-L18	Cyanobacteria bacterium
14	MSK.17.78	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Photobacterium damselae	Opitutae bacterium SCGC AG-212-L18	Cyanobacteria bacterium
15	MSK.17.79	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Cyanobacteria bacterium	Photobacterium damselae	Opitutae bacterium SCGC AG-212-L18
16	MSK.22.23	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Cyanobacteria bacterium	Photobacterium damselae	Candidatus Tremblaya princeps
17	MSK.22.28	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Cyanobacteria bacterium	Photobacterium damselae	Candidatus Tremblaya princeps
18	MSK.22.51	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Cyanobacteria bacterium	Photobacterium damselae	Candidatus Tremblaya princeps
19	MSK.16.45	[Eubacterium] rectale	[Eubacterium] rectale	Cyanobacteria bacterium	Candidatus Vidania fulgoroideae	Candidatus Carsonella ruddii	Bacillus sp. UFRGS-B20
20	MSK.22.92	[Eubacterium] rectale	[Eubacterium] rectale	Roseburia sp. CAG:18	Cyanobacteria bacterium	Photobacterium damselae	Candidatus Vidania fulgoroideae

Conservation of motility expanding out from E. rectale

# for each node find all children and measure percentage of children where 
# chosen oggs are present
cladesateachtreelevel = getleafnames.(speciesparents);
frac_present = map(cladesateachtreelevel) do clade_ids
    cladeidx = indexin(clade_ids, uniprot.obs_names)
    mean(subset_ogg_mtx[cladeidx, :], dims=1) |> vec
end |> stack

# check how many oggs (n=16) are fully conserved across the smallest clade (n=3)
islocallypresent = frac_present[:, 1] .== 1;
@info "# locally present = $(sum(islocallypresent)), # not locally present = $(sum(.!islocallypresent))"
numcols = length(cladesizes);

┌ Info: # locally present = 12, # not locally present = 4
└ @ Main /Users/bend/projects/Doran_etal_2023/notebooks/jl_notebook_cell_df34fa98e69747e1a8f8a730347b8e2f_Y120sZmlsZQ==.jl:11

# plot the locally present oggs up through the tree
plot(
    xticks=(1:numcols, string.(cladesizes)),
    xrotation=45,
    ylabel="fractional coverage of taxa",
    xlabel="# taxa in clade",
    margin=5Plots.Measures.mm,
    ylims=(0,1), widen=true,
    size=(600,300),
    yticks=0:.25:1,
    legend=:bottomleft,
)
hline!([.25, .5, .75], c=:grey, alpha=.5, linestyle=:dash)
violin!(permutedims(1:numcols), frac_present[islocallypresent, :], alpha=.3, c="#6182ce")
dotplot!(permutedims(1:numcols), frac_present[islocallypresent, :], c="#6182ce", alpha=.5, mode=:none, )
scatter!([NaN], label=permutedims(["$(species_name) locally conserved OGG (n=$(sum(islocallypresent)))"]), c=["#6182ce"], alpha=[0.5] )

these results seemed to show that indeed the orthologs absent from phage-assoiated strains in E. rectale are present on over half of the 250 most closely related species. And, even across all bacteria, there at least 3 orthologs that are conserved across half of bacteria.

What phage related differences are present for other species?

Given the results in E. rectale we wanted to explore if these results were consistent across other species.

function plot_conserved_oggs(species_title, chosen_oggs)
    chosen_oggs_idx = indexin(chosen_oggs, uniprot.var_names);
    coverage_uniprot = mean(upmtx .> 0, dims=1) |> vec
    sample_coverage = coverage_uniprot# [chosen_oggs_idx]
    sample_percentile = map(sample_coverage) do x 
        mean(coverage_uniprot .< x)
    end
    order = sortperm(sample_percentile)
    plot(
        title=species_title,
        xlabel="percentile",
        ylabel="fractional presence",
        lims=(0,1), 
        # widen=true, 
        ratio=1,
        grid=true,
        ticks=[0, .5, 1],
        tickdirection=:out,
        framestyle=:box,
    )


    plot!(sample_percentile[order], sample_coverage[order],
        c=:lightgrey, linewidth=4,
    )
    sample_coverage = coverage_uniprot[chosen_oggs_idx]
    sample_percentile = map(sample_coverage) do x 
        mean(coverage_uniprot .< x)
    end
    p1 = scatter!(sample_percentile, sample_coverage,
        alpha=.5,c="#6182ce", markersize=6,
    )
    margindensity = density(coverage_uniprot, 
        trim=true, fill=0, c=:yellow, linecolor=:black, permute=(:x, :y),
        ylabel="density",  yrotation=270, yticks=0:5:10,
        # xticks=false,
    )
    plot(p1, margindensity, layout = @layout([a{.9w} b]), size=(500,400), link=:y)
end

plot_conserved_oggs (generic function with 1 method)

function getuniquedonors_fromtree(tree, target_id) 
    node = filter!(n -> id(n) == target_id, prewalk(tree))[1]
    grp1 = getleafnames(children(node)[1])
    doner_ids_1 = join(unique(replace.(grp1, r"\.\d+$"=>"")), ";")
    grp2 = getleafnames(children(node)[2])
    doner_ids_2 = join(unique(replace.(grp2, r"\.\d+$"=>"")), ";")
    doner_ids_1, doner_ids_2
end
function getuniquestrains_fromtree(tree, target_id) 
    node = filter!(n -> id(n) == target_id, prewalk(tree))[1]
    grp1 = getleafnames(children(node)[1])
    doner_ids_1 = join(grp1, ";")
    grp2 = getleafnames(children(node)[2])
    doner_ids_2 = join(grp2, ";")
    doner_ids_1, doner_ids_2
end

getuniquestrains_fromtree (generic function with 1 method)

strain_level_testresults = testresults |>
    df->subset(df, 
        :tree_level => ByRow(==("strain level")), 
        :log2FC => ByRow(x-> abs(x) > 1), 
        :qval_BH => ByRow(<(0.05)),
        ) |>
    df->leftjoin(df, uniprot.var[:, [:og, :description]], on = :ogg_name => :og) |>
    df->subset(df, :description => ByRow(x-> (!ismissing(x)) && (x!=("")))) |>
    df->select(df, :nodeids, :nodedepth, :grp1_species_mode => :species, :grp1_N, :grp2_N, :ogg_name,:ogg_freqs, :grp1_prp_exp, :grp2_prp_exp, :log2FC, :qval_BH, :description) |>
    df->groupby(df, :nodeids) |>
    df->transform(df, 
        :nodeids => ByRow(x -> getuniquedonors_fromtree(strvartree, x)) => [:grp1_donor_ids, :grp2_donor_ids],
        :nodeids => ByRow(x -> getuniquestrains_fromtree(strvartree, x)) => [:grp1_strain_ids, :grp2_strain_ids],
    ) |>
    df->sort(df, [:nodeids, :log2FC])
first(strain_level_testresults, 5)

5×16 DataFrame

Row	nodeids	nodedepth	species	grp1_N	grp2_N	ogg_name	ogg_freqs	grp1_prp_exp	grp2_prp_exp	log2FC	qval_BH	description	grp1_donor_ids	grp2_donor_ids	grp1_strain_ids	grp2_strain_ids
	Int64	Float64	String31	Int64	Int64	String15	Float64	Float64	Float64	Float64	Float64	String?	String	String	String	String
1	4	1.59622	Bacteroides uniformis	4	23	COG1115	0.924157	0.25	1.0	-1.3891	0.0	amino acid carrier protein	DFI.1;MSK.17	MSK.19;MSK.20;MSK.16;DFI.3;MSK.18;DFI.4	DFI.1.247;DFI.1.135;DFI.1.173;MSK.17.15	MSK.19.71;MSK.19.54;MSK.19.4;MSK.19.91;MSK.19.84;MSK.20.30;MSK.20.3;MSK.20.79;MSK.16.71;MSK.16.7;MSK.16.46;MSK.16.44;MSK.16.39;DFI.3.57;MSK.18.66;MSK.18.33;MSK.18.78;MSK.18.89;MSK.18.30;MSK.18.25;MSK.18.77;MSK.18.74;DFI.4.149
2	4	1.59622	Bacteroides uniformis	4	23	COG0639	0.266854	0.25	0.956522	-1.26605	0.0	phosphoprotein phosphatase activity	DFI.1;MSK.17	MSK.19;MSK.20;MSK.16;DFI.3;MSK.18;DFI.4	DFI.1.247;DFI.1.135;DFI.1.173;MSK.17.15	MSK.19.71;MSK.19.54;MSK.19.4;MSK.19.91;MSK.19.84;MSK.20.30;MSK.20.3;MSK.20.79;MSK.16.71;MSK.16.7;MSK.16.46;MSK.16.44;MSK.16.39;DFI.3.57;MSK.18.66;MSK.18.33;MSK.18.78;MSK.18.89;MSK.18.30;MSK.18.25;MSK.18.77;MSK.18.74;DFI.4.149
3	4	1.59622	Bacteroides uniformis	4	23	COG3958	0.842697	0.0	1.0	-1.12717	0.0	transketolase activity	DFI.1;MSK.17	MSK.19;MSK.20;MSK.16;DFI.3;MSK.18;DFI.4	DFI.1.247;DFI.1.135;DFI.1.173;MSK.17.15	MSK.19.71;MSK.19.54;MSK.19.4;MSK.19.91;MSK.19.84;MSK.20.30;MSK.20.3;MSK.20.79;MSK.16.71;MSK.16.7;MSK.16.46;MSK.16.44;MSK.16.39;DFI.3.57;MSK.18.66;MSK.18.33;MSK.18.78;MSK.18.89;MSK.18.30;MSK.18.25;MSK.18.77;MSK.18.74;DFI.4.149
4	4	1.59622	Bacteroides uniformis	4	23	COG3959	0.842697	0.0	1.0	-1.12717	0.0	transketolase activity	DFI.1;MSK.17	MSK.19;MSK.20;MSK.16;DFI.3;MSK.18;DFI.4	DFI.1.247;DFI.1.135;DFI.1.173;MSK.17.15	MSK.19.71;MSK.19.54;MSK.19.4;MSK.19.91;MSK.19.84;MSK.20.30;MSK.20.3;MSK.20.79;MSK.16.71;MSK.16.7;MSK.16.46;MSK.16.44;MSK.16.39;DFI.3.57;MSK.18.66;MSK.18.33;MSK.18.78;MSK.18.89;MSK.18.30;MSK.18.25;MSK.18.77;MSK.18.74;DFI.4.149
5	4	1.59622	Bacteroides uniformis	4	23	COG1226	0.241573	1.0	1.0	-1.06998	0.0	(belongs to the monovalent cation proton antiporter 2 (CPA2) transporter (TC 2.A.37) family)	DFI.1;MSK.17	MSK.19;MSK.20;MSK.16;DFI.3;MSK.18;DFI.4	DFI.1.247;DFI.1.135;DFI.1.173;MSK.17.15	MSK.19.71;MSK.19.54;MSK.19.4;MSK.19.91;MSK.19.84;MSK.20.30;MSK.20.3;MSK.20.79;MSK.16.71;MSK.16.7;MSK.16.46;MSK.16.44;MSK.16.39;DFI.3.57;MSK.18.66;MSK.18.33;MSK.18.78;MSK.18.89;MSK.18.30;MSK.18.25;MSK.18.77;MSK.18.74;DFI.4.149

CSV.write(joinpath(rdir, "strain_level_testresults_with_annotations.tsv"), strain_level_testresults; delim="\t")

"/Users/bend/projects/Doran_etal_2023/_research/strain_variation_wilcoxon_tests/strain_level_testresults_with_annotations.tsv"

uniqueOGGresults |>
    df->subset(df, 
        :tree_level => ByRow(==("strain level")), 
        :log2FC => ByRow(x-> abs(x) > 1), 
        :qval_BH => ByRow(<(0.05)),
        ) |>
    df->leftjoin(df, uniprot.var[:, [:og, :description]], on = :ogg_name => :og) |>
    df->subset(df, :description => ByRow(x-> (!ismissing(x)) && (x!=("")))) |>
    df->select(df, :nodeids, :nodedepth, :grp1_species_mode => :species, :grp1_N, :grp2_N, :ogg_name,:ogg_freqs, :grp1_prp_exp, :grp2_prp_exp, :log2FC, :qval_BH, :description) |>
    df->groupby(df, :nodeids) |>
    df->transform(df, 
        :nodeids => ByRow(x -> getuniquedonors_fromtree(strvartree, x)) => [:grp1_donor_ids, :grp2_donor_ids],
        :nodeids => ByRow(x -> getuniquestrains_fromtree(strvartree, x)) => [:grp1_strain_ids, :grp2_strain_ids],
    ) |>
    df->sort(df, [:nodeids, :log2FC]) |>
    df->CSV.write(joinpath(rdir, "uniqueOGGresults_strainlevel_testresults.tsv"), df; delim="\t")

"/Users/bend/projects/Doran_etal_2023/_research/strain_variation_wilcoxon_tests/uniqueOGGresults_strainlevel_testresults.tsv"

Exploring these tables of test results revealed other instances of phage related strain level differences. For P. vulgatus, M. gnavus, B. thetaiotaomicron, and C. comes we found cases where OGGs associated with phage infection where more abundant in multiple strains in contrast to other “phage suppressed OGGs”. We found no specificity in degree of conservation for these phage suppressed OGGs, highly conserved OGGs were just as likely to be suppressed as OGGs only found in a small number of taxa across all of UniProt.

P. vulgatus

species_title = "Phocaeicola vulgatus"
targetid = "DFI.3.23"
strvar_leaves = getleaves(strvartree)
basenode = strvar_leaves[findfirst(n->name(n) == targetid, strvar_leaves)]
subtree = readnw(nwstr(nthparent(basenode, 4)))
ladderize!(subtree)
# tiplabels = join.(eachrow(strvarobs[indexin(getleafnames(subtree), strvarobs.Strain_ID), [:Strain_ID, :NCBI_Species]]), " ")
# idmapping = Dict(k=>v for (k,v) in zip(getleafnames(subtree), tiplabels))
# rename_treeleaves!(subtree, idmapping)

plot(subtree,
    c=:black,
    lw=3, fs=5,
    size=(600,600),
    rightmargin=3Plots.cm,
    leftmargin=10Plots.mm,
    ylabel=species_title * " (n = $(length(getleaves(subtree))))",
    framestyle=:grid,
    ticks=false,
)

# find and plot test nodes in subtree
leaves_of_test_nodes = map(prewalk(nthparent(basenode, 4))) do node; test_nodes=unique(strain_level_testresults.nodeids)
    if (Int(id(node)) ∈ test_nodes) && minimum(length.(getleaves.(children(node)))) > 6
        join(sort(getleafnames(node)), ";")
    else
        nothing
    end
end |> x->filter(!isnothing, x)
subtreepos = NewickTree.treepositions(subtree)
testnode_positions = map(keys(subtreepos), values(subtreepos)) do k,v
    x, y = v
    label = join(sort(getleafnames(k)), ";")
    if label ∈ leaves_of_test_nodes
        (; x, y, label)
    else
        nothing
    end
end |> x->filter(!isnothing, x)
@df DataFrame(testnode_positions) scatter!(:x, :y, markersize=6, c=:yellow, markershape=:diamond)

# collected by manually searching through strain-level test results
notable_oggs = ["COG0863","COG1083","COG4744", "32ZV1", "COG0666","COG2214","2Z8QV"]
plot_conserved_oggs("P. vulgatus", notable_oggs) 

M. gnavus

species_title = "[Ruminococcus] gnavus"
targetid = "MSK.22.91"
strvar_leaves = getleaves(strvartree)
basenode = strvar_leaves[findfirst(n->name(n) == targetid, strvar_leaves)]
subtreeroot = nthparent(basenode, 5)
subtree = readnw(nwstr(subtreeroot))
ladderize!(subtree)
# tiplabels = join.(eachrow(strvarobs[indexin(getleafnames(subtree), strvarobs.Strain_ID), [:Strain_ID, :NCBI_Species]]), " ")
# idmapping = Dict(k=>v for (k,v) in zip(getleafnames(subtree), tiplabels))
# rename_treeleaves!(subtree, idmapping)

plot(subtree,
    c=:black,
    lw=3, fs=5,
    size=(600,600),
    rightmargin=3Plots.cm,
    leftmargin=10Plots.mm,
    ylabel="M. gnavus (n = 41)",
    framestyle=:grid,
    ticks=false,
)

# find and plot test nodes in subtree
leaves_of_test_nodes = map(prewalk(subtreeroot)) do node; test_nodes=unique(strain_level_testresults.nodeids)
    if (Int(id(node)) ∈ test_nodes) && minimum(length.(getleaves.(children(node)))) >= 4
        join(sort(getleafnames(node)), ";")
    else
        nothing
    end
end |> x->filter(!isnothing, x)
subtreepos = NewickTree.treepositions(subtree)
testnode_positions = map(keys(subtreepos), values(subtreepos)) do k,v
    x, y = v
    label = join(sort(getleafnames(k)), ";")
    if label ∈ leaves_of_test_nodes
        (; x, y, label)
    else
        nothing
    end
end |> x->filter(!isnothing, x)
@df DataFrame(testnode_positions) scatter!(:x, :y, markersize=6, c=:yellow, markershape=:diamond)

# collected by manually searching through strain-level test results
notable_oggs =     ["COG1440","COG1447","COG1694", "COG0531","COG1501","COG2222"]
plot_conserved_oggs("M. gnavus", notable_oggs) 

B. thetaiotaomicron

species_title = "Bacteroides thetaiotaomicron"
targetid = "DFI.4.108"
strvar_leaves = getleaves(strvartree)
basenode = strvar_leaves[findfirst(n->name(n) == targetid, strvar_leaves)]
subtreeroot = nthparent(basenode, 4)
subtree = readnw(nwstr(subtreeroot))
ladderize!(subtree)
# tiplabels = join.(eachrow(strvarobs[indexin(getleafnames(subtree), strvarobs.Strain_ID), [:Strain_ID, :NCBI_Species]]), " ")
# idmapping = Dict(k=>v for (k,v) in zip(getleafnames(subtree), tiplabels))
# rename_treeleaves!(subtree, idmapping)

plot(subtree,
    c=:black,
    lw=3, fs=5,
    size=(600,600),
    rightmargin=3Plots.cm,
    leftmargin=10Plots.mm,
    ylabel=species_title * " (n = $(length(getleaves(subtree))))",
    framestyle=:grid,
    ticks=false,
)

# find and plot phage related test nodes in subtree
leaves_of_test_nodes = map(prewalk(subtreeroot)) do node; test_nodes=unique(strain_level_testresults.nodeids)
    if (Int(id(node)) ∈ test_nodes) && minimum(length.(getleaves.(children(node)))) == 7
        join(sort(getleafnames(node)), ";")
    else
        nothing
    end
end |> x->filter(!isnothing, x)
subtreepos = NewickTree.treepositions(subtree)
testnode_positions = map(keys(subtreepos), values(subtreepos)) do k,v
    x, y = v
    label = join(sort(getleafnames(k)), ";")
    if label ∈ leaves_of_test_nodes
        (; x, y, label)
    else
        nothing
    end
end |> x->filter(!isnothing, x)
@df DataFrame(testnode_positions) scatter!(:x, :y, markersize=6, c=:yellow, markershape=:diamond)

# collected by manually searching through strain-level test results
notable_oggs = ["COG1474","COG2942","COG3410"]
plot_conserved_oggs("B. thetaiotaomicron", notable_oggs)

C. comes

species_title = "Coprococcus comes"
targetid = "MSK.20.88"
strvar_leaves = getleaves(strvartree)
basenode = strvar_leaves[findfirst(n->name(n) == targetid, strvar_leaves)]
subtreeroot = nthparent(basenode, 6)
subtree = readnw(nwstr(subtreeroot))
ladderize!(subtree)
# tiplabels = join.(eachrow(strvarobs[indexin(getleafnames(subtree), strvarobs.Strain_ID), [:Strain_ID, :NCBI_Species]]), " ")
# idmapping = Dict(k=>v for (k,v) in zip(getleafnames(subtree), tiplabels))
# rename_treeleaves!(subtree, idmapping)

plot(subtree,
    c=:black,
    lw=3, fs=5,
    size=(600,600),
    rightmargin=3Plots.cm,
    leftmargin=10Plots.mm,
    ylabel=species_title * " (n = $(length(getleaves(subtree))))",
    framestyle=:grid,
    ticks=false,
)

# find and plot phage related test nodes in subtree
leaves_of_test_nodes = map(prewalk(subtreeroot)) do node; test_nodes=unique(strain_level_testresults.nodeids)
    if (Int(id(node)) ∈ test_nodes) && minimum(length.(getleaves.(children(node)))) ==5
        join(sort(getleafnames(node)), ";")
    else
        nothing
    end
end |> x->filter(!isnothing, x)
subtreepos = NewickTree.treepositions(subtree)
testnode_positions = map(keys(subtreepos), values(subtreepos)) do k,v
    x, y = v
    label = join(sort(getleafnames(k)), ";")
    if label ∈ leaves_of_test_nodes
        (; x, y, label)
    else
        nothing
    end
end |> x->filter(!isnothing, x)
@df DataFrame(testnode_positions) scatter!(:x, :y, markersize=6, c=:yellow, markershape=:diamond)

# collected by manually searching through strain-level test results
notable_oggs = ["COG0145","COG1011","COG1351","COG4951","2Z81G",]
plot_conserved_oggs("C. comes", notable_oggs)