Truvari Metrics PairPlot

[ACEnglish]

With the addition of FP/FN bench annotation, a seaborn PairPlot is now super informative. Deeper exploration of using the new bench annotations is described here.

Below, I'll also show how to make a PairPlot for a quick overview of the Truvari metrics (Figure 3A). Zooming into one pair of metrics and adding svtype markers shows even more detail (Figure 3B). Note we don't see any blue 'tpbase' dots because --multimatch wasn't used and 'tpbase' metrics are identical to 'tp'.

import joblib
import pandas as pd
import seaborn as sb
import matplotlib.pyplot as plt

v310_data = joblib.load("results_31new_seq/data.jl")

# We'll put TruScore on a 0 to 1 Scale so that our diagonal histogram have comparable bins
v310_data["TruScore_"] = v310_data["TruScore"] / 100

# These metrics we wish to plot
metrics = ["PctSeqSimilarity", "PctSizeSimilarity", "PctRecOverlap", "TruScore_"]

# Make plot
g = sb.PairGrid(data=v310_data.reset_index(), vars=metrics, hue="state")
g.map_diag(sb.histplot, binwidth=0.1)
g.map_offdiag(sb.scatterplot)
g.add_legend()
plt.suptitle('Fig3A - Truvari Metrics',  y=1.01, 
             size = 24);
plt.show()

plt.figure(dpi=180)
p = sb.scatterplot(data=v310_data.reset_index(), 
                   x="TruScore", 
                   y="PctSeqSimilarity", 
                   hue="state", 
                   style="svtype", 
                   style_order=["DEL", "INS"])
p.set(title="Fig3B")
sb.move_legend(p, "upper left", bbox_to_anchor=(1, 1))