#! /usr/bin/env julia
# Installing Dependencies
## using Pkg
## Pkg.add("Distributed")
## Pkg.add("DataFrames")
## Pkg.add("CSV")
## Pkg.add("SNaQ")
## Pkg.add("PhyloNetworks")
## Pkg.add("RCall")
## Pkg.add("PhyloPlots")
## Pkg.add("QuartetNetworkGoodnessFit")

# Running SNaQ Analysis
using PhyloNetworks, SNaQ;
using Distributed;
addprocs(9);
@everywhere using PhyloNetworks, SNaQ;
nruns = 100; # number of runs for each hmax
astralfile = joinpath("..", "species_tree", "coal.tre");
astraltree = readnewick(astralfile);

### Reading RAxML gene trees and ASTRAL species tree
### running in raxml_snaq/ folder
# raxmltrees = joinpath("..", "..", "species_tree", "all.trees");
# inputCF = readtrees2CF(raxmltrees);
# net0 = snaq!(astraltree, inputCF, hmax=0, filename="net0", seed=123, outgroup="ZJ", runs=nruns);
# net1 = snaq!(net0, inputCF, hmax=1, filename="net1", seed=123, outgroup="ZJ", runs=nruns);
# net2 = snaq!(net1, inputCF, hmax=2, filename="net2", seed=123, outgroup="ZJ", runs=nruns);
# net3 = snaq!(net2, inputCF, hmax=3, filename="net3", seed=123, outgroup="ZJ", runs=nruns);
# net4 = snaq!(net3, inputCF, hmax=4, filename="net4", seed=123, outgroup="ZJ", runs=nruns);

### Alternatively, reading in the input files from Bucky
### running in input_snaq/ folder
inputCFfile = joinpath("..", "bucky", "bucky_1", "input.mb.CFs.csv");
inputCF = readtableCF(inputCFfile);
net0 = snaq!(astraltree, inputCF, hmax=0, filename="net0", seed=123, outgroup="ZJ", runs=nruns);
net1 = snaq!(net0, inputCF, hmax=1, filename="net1", seed=123, outgroup="ZJ", runs=nruns);
net2 = snaq!(net1, inputCF, hmax=2, filename="net2", seed=123, outgroup="ZJ", runs=nruns);
net3 = snaq!(net2, inputCF, hmax=3, filename="net3", seed=123, outgroup="ZJ", runs=nruns);
net4 = snaq!(net3, inputCF, hmax=4, filename="net4", seed=12345, outgroup="ZJ", runs=nruns);


# Plotting the SNaQ results
using PhyloPlots, RCall;
using PhyloNetworks, SNaQ;
## Network scores vs. hmax
net0 = readsnaqnetwork("net0.out");
net1 = readsnaqnetwork("net1.out");
net2 = readsnaqnetwork("net2.out");
net3 = readsnaqnetwork("net3.out");
net4 = readsnaqnetwork("net4.out");
scores = [loglik(net0), loglik(net1), loglik(net2), loglik(net3), loglik(net4)];
hmax = collect(0:4);
R"pdf"("snaq_network_scores.pdf", width=12, height=8);
R"plot"(hmax, scores, type="b", ylab="network score", xlab="hmax", col="blue");
R"dev.off"();

## Rerooting and rotating the networks for better visualization
rootatnode!(net1, "ZJ");
rootatnode!(net2, "ZJ");
rootatnode!(net3, "ZJ");
rootatnode!(net4, "ZJ");
### rotate!(net1, -2);
### rotate!(net2, -2);
### rotate!(net3, -2);
### rotate!(net4, -2);

## Plotting the networks
R"pdf"("snaq_networks_fulltree.pdf", width=14, height=10);
R"layout(matrix(1:4, 2, 2, byrow=TRUE))"; # to get 4 plots into a single figure: 2 row, 2 columns
R"par"(mar=[0, 0, 1.5, 0]);        # for smaller margins
xmin, xmax = PhyloPlots.PhyloPlots.edgenode_coordinates(net1, false, false)[13:14];
xmax += (xmax - xmin) * 0.3;
plot(net1, showgamma=true, tipoffset=0.1, xlim=[xmin, xmax]);
R"mtext"(string("hmax=1, loglik=-", round(loglik(net1), digits=2)), font=2);
xmin, xmax = PhyloPlots.PhyloPlots.edgenode_coordinates(net2, false, false)[13:14];
xmax += (xmax - xmin) * 0.3;
plot(net2, showgamma=true, tipoffset=0.1, xlim=[xmin, xmax]);
R"mtext"(string("hmax=2, loglik=-", round(loglik(net2), digits=2)), font=2);
xmin, xmax = PhyloPlots.PhyloPlots.edgenode_coordinates(net3, false, false)[13:14];
xmax += (xmax - xmin) * 0.3;
plot(net3, showgamma=true, tipoffset=0.1, xlim=[xmin, xmax]);
R"mtext"(string("hmax=3, loglik=-", round(loglik(net3), digits=2)), font=2);
xmin, xmax = PhyloPlots.PhyloPlots.edgenode_coordinates(net4, false, false)[13:14];
xmax += (xmax - xmin) * 0.3;
plot(net4, showgamma=true, tipoffset=0.1, xlim=[xmin, xmax]);
R"mtext"(string("hmax=4, loglik=-", round(loglik(net4), digits=2)), font=2);
R"dev.off"();

R"pdf"("snaq_networks_majortree.pdf", width=14, height=10);
R"layout(matrix(1:4, 2, 2, byrow=TRUE))"; # to get 4 plots into a single figure: 2 row, 2 columns
R"par"(mar=[0, 0, 1.5, 0]);        # for smaller margins
xmin, xmax = PhyloPlots.PhyloPlots.edgenode_coordinates(net1, false, false)[13:14];
xmax += (xmax - xmin) * 0.3;
plot(net1, style=:majortree, showgamma=true, tipoffset=0.1, xlim=[xmin, xmax]);
R"mtext"(string("hmax=1, loglik=-", round(loglik(net1), digits=2)), font=2);
xmin, xmax = PhyloPlots.PhyloPlots.edgenode_coordinates(net2, false, false)[13:14];
xmax += (xmax - xmin) * 0.3;
plot(net2, style=:majortree, showgamma=true, tipoffset=0.1, xlim=[xmin, xmax]);
R"mtext"(string("hmax=2, loglik=-", round(loglik(net2), digits=2)), font=2);
xmin, xmax = PhyloPlots.PhyloPlots.edgenode_coordinates(net3, false, false)[13:14];
xmax += (xmax - xmin) * 0.3;
plot(net3, style=:majortree, showgamma=true, tipoffset=0.1, xlim=[xmin, xmax]);
R"mtext"(string("hmax=3, loglik=-", round(loglik(net3), digits=2)), font=2);
xmin, xmax = PhyloPlots.PhyloPlots.edgenode_coordinates(net4, false, false)[13:14];
xmax += (xmax - xmin) * 0.3;
plot(net4, style=:majortree, showgamma=true, tipoffset=0.1, xlim=[xmin, xmax]);
R"mtext"(string("hmax=4, loglik=-", round(loglik(net4), digits=2)), font=2);
R"dev.off"();

## expected vs. observed quartet concordance factors
using CSV, DataFrames, Distributions, Random, RCall;
using PhyloNetworks, SNaQ;
using QuartetNetworkGoodnessFit;
using Statistics, Printf;

inputCFfile = joinpath("..", "bucky", "bucky_1", "input.mb.CFs.csv");
inputCF = readtableCF(inputCFfile);
Random.seed!(123);
jitter = 0.005;
hmax_list = 0:3;

R"pdf"("expCFs_obsvsfitted.pdf", width=6, height=6);
R"par(mfrow=c(2,2), mar=c(4,4,3,1))"
gof_results = DataFrame(h=Int[], pvalue_normal=Float64[], pvalue_empirical=Float64[], z_uncorrected=Float64[], sigma=Float64[], nsim=Int[])

for h in hmax_list
    net = readsnaqnetwork("net$(h).out")
    topologymaxQpseudolik!(net, inputCF)

    df_long = fittedquartetCF(inputCF, :long)
    df_jittered = deepcopy(df_long)
    df_jittered.obsCF .+= rand(Uniform(-jitter/2, jitter/2), nrow(df_long))
    df_jittered.expCF .+= rand(Uniform(-jitter/2, jitter/2), nrow(df_long))

    res = quarnetGoFtest!(net, inputCF, false; seed=123, nsim=1000)
    z_obs = res[2]
    z_sim = sort!(res[6])
    p_emp = mean(z_sim .>= z_obs)      # 经验 p 值：模拟 z 超过观测 z 的比例

    # 存表时两种 p 值都留，加上模拟分布的特征备查
    push!(gof_results, (h=h,
        pvalue_normal=res[1],
        pvalue_empirical=p_emp,
        z_uncorrected=z_obs,
        sigma=res[3],
        nsim=1000))


    ptxt = p_emp == 0 ? "p < 0.001 (emp.)" : @sprintf("p = %.3f (emp.)", p_emp)
    ttl = "h = $h, $ptxt"
    R"plot"(df_jittered.obsCF, df_jittered.expCF,
        xlim=[0, 1.05], ylim=[0, 1.05],
        xlab="quartet CF observed in gene trees",
        ylab="quartet CF expected from network",
        main=ttl,
        pch=16, col="#008080", cex=0.5)
    R"abline"(0, 1, col="#008080", lwd=0.5)
end

R"dev.off"();

CSV.write("gof_results.csv", gof_results)
println(gof_results)


net_final = readsnaqnetwork("net0.out");
topologymaxQpseudolik!(net_final, inputCF);
df_long = fittedquartetCF(inputCF, :long);

### Adding jitter to the points for better visualization
df_jittered = deepcopy(df_long);
df_jittered.obsCF .+= rand(Uniform(-jitter / 2, jitter / 2), nrow(df_long));
df_jittered.expCF .+= rand(Uniform(-jitter / 2, jitter / 2), nrow(df_long));
df_wide = fittedquartetCF(inputCF, :wide)
println(names(df_wide))
df_wide.outlier_p = res1[4]
outliers = sort(df_wide[df_wide.outlier_p .< 0.01, :], :outlier_p)
CSV.write("outlier_quartets_h1.csv", outliers)


R"plot"(df_jittered.obsCF, df_jittered.expCF, xlab="quartet CF observed in gene trees",
    ylab="quartet CF expected from network", pch=16,
    col="#008080",
    cex=0.7);
R"abline"(0, 1, col="#008080", lwd=0.5);
R"dev.off"();

# Goodness of fit of the SNaQ networks
using QuartetNetworkGoodnessFit;
res1 = quarnetGoFtest!(net_final, inputCF, true; seed=123, nsim=1000);
res1[[1, 2, 3]] # p-value, uncorrected z, σ
z_obs = res1[2]
z_sim = sort!(res1[6])
p_emp = mean(z_sim .>= z_obs)