Br analysis scripts

analysis/fig-s4.py

@@ -0,0 +1,94 @@
+import pandas as pd
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+import tqdm
+import colorcet
+
+sns.set(font_scale=1.7)
+sns.set_style("whitegrid")
+plt.rcParams["axes.grid"] = False
+plt.rc("axes", edgecolor="black")
+plt.rc(
+    "text.latex",
+    preamble=r"\usepackage{newpxtext}\usepackage{newpxmath}\usepackage{commath}\usepackage{mathtools}",
+)
+plt.rc("font", family="serif", size=16.0, weight="medium")
+plt.rc("savefig", dpi=500)
+plt.rc("legend", loc="best", fontsize="medium", fancybox=True, framealpha=0.5)
+plt.rc("lines", linewidth=2.5, markersize=10, markeredgewidth=2.5)
+plt.rc("axes", titlepad=10)
+
+
+both = pd.read_csv('pdb_plus_240k/scRMSD_best_240k_plus_pdbFirst_aln0_cluster.tsv', sep='\t',
+                names=['representative', 'member'])
+
+output_rows = []
+for resample_freq in tqdm.tqdm((0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9)):
+    for iter_n in range(5):
+        resampled = both.sample(frac=resample_freq, random_state=iter_n, replace=False)
+        n_uniq_clust = resampled['representative'].nunique()
+        output_rows.append(dict(
+            freq=resample_freq,
+            iteration=iter_n,
+            num_uniq_clust=n_uniq_clust,
+            comparison=name,
+        ))
+
+    output_rows.append(dict(freq=1, iteration=0, comparison=name, num_uniq_clust=both.representative.nunique()))
+    output_rows.append(dict(freq=0, iteration=0, comparison=name, num_uniq_clust=0))
+
+df_full_resamp = pd.DataFrame(output_rows)
+
+output_rows = []
+for full_df, name in zip((both,), ('PDB + syn',)):
+    for resample_freq in tqdm.tqdm((0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9)):
+        # names of synthetic samples will start with date
+        rest_df = full_df[~full_df['member'].str.startswith("2024")]
+        df = full_df[full_df['member'].str.startswith("2024")]
+        rest_u = rest_df.representative.nunique()
+
+        for iter_n in range(5):
+            resampled = df.sample(frac=resample_freq, random_state=iter_n, replace=False)
+            n_uniq_clust = resampled['representative'].nunique() + rest_u
+            output_rows.append(dict(
+                freq=resample_freq,
+                iteration=iter_n,
+                num_uniq_clust=n_uniq_clust,
+                comparison=name,
+            ))
+
+    output_rows.append(dict(freq=1, iteration=0, comparison=name, num_uniq_clust=full_df.representative.nunique()))
+    output_rows.append(dict(freq=0, iteration=0, comparison=name, num_uniq_clust=0))
+
+df_partial_resamp = pd.DataFrame(output_rows)
+
+plt.figure()
+sns.lineplot(data=df_partial_resamp, 
+            x='freq', 
+            y='num_uniq_clust', 
+            label='Partial resampling',
+            color='#0D96C9',)
+
+# Plot the full resampling line in blue
+sns.lineplot(data=df_full_resamp, 
+            x='freq', 
+            y='num_uniq_clust', 
+            label='Full resampling',
+             color='#404040',
+            )
+
+plt.xlabel("Proportion of data points \nsampled")
+plt.ylabel("No. of clusters")
+plt.xticks(np.arange(0, 1.1, 0.1), rotation=75)
+plt.legend(title=None, loc='best', frameon=False)
+plt.title("Number of distinct clusters by\n resampling frequency")
+
+# Remove top and right spines
+plt.gca().spines['top'].set_visible(False)
+plt.gca().spines['right'].set_visible(False)
+
+plt.tight_layout()
+
+
+

analysis/fig-s5.py

@@ -0,0 +1,89 @@
+#!/usr/bin/env python
+# coding: utf-8
+
+# In[1]:
+
+
+import pathlib
+import pandas as pd
+import pathlib
+import pandas as pd
+import joblib
+import numpy as np
+import seaborn as sns
+import matplotlib.pyplot as plt
+
+import re
+
+sns.set(font_scale=1.7)
+sns.set_style("whitegrid")
+plt.rcParams['axes.grid'] = False
+plt.rc('axes',edgecolor='black')
+
+plt.rc("text", usetex=False)
+plt.rc(
+    "text.latex",
+    preamble=r"\usepackage{newpxtext}\usepackage{newpxmath}\usepackage{commath}\usepackage{mathtools}",
+)
+plt.rcParams['font.family'] = 'sans-serif'
+plt.rcParams['font.sans-serif'] = ['DejaVu Sans']
+# plt.rc("font", family="serif", size=16.0, weight="medium")
+plt.rc("savefig", dpi=500)
+plt.rc("legend", loc="best", fontsize="medium", fancybox=True, framealpha=0.5)
+plt.rc("lines", linewidth=2.5, markersize=10, markeredgewidth=2.5)
+plt.rc("axes", titlepad=10)
+
+colors = ["#BBBBBB", "#33BBEE", "#EE3377", "#009988", "#CC3311", "#0077BB"]
+colors = list(reversed(colors))
+sns.set_palette(sns.color_palette(colors))
+# set mpl palette
+plt.rcParams["axes.prop_cycle"] = plt.cycler(color=colors)
+
+
+# In[2]:
+
+
+df = pd.read_parquet('temperature_scRMSD_gridsearch_results.parquet')
+
+
+# In[3]:
+
+
+averages = df.groupby(["temperature", 
+                      'backbone_pdb'])[
+    ["scRMSD", "TM", "aa_length"]
+].mean()
+averages = averages.reset_index()
+
+averages["pass"] = averages["scRMSD"] <= 2
+averages["pass_tm"] = averages["TM"] >= 0.5
+bins = np.arange(0, 1100, 100)
+
+
+# In[4]:
+
+
+plt.figure() 
+averages['pass_pct'] = averages['pass'] * 100
+sns.lineplot(data=averages, 
+           x='temperature', 
+           y='pass_pct',
+           marker='o',
+           color='#0D96C9', 
+           errorbar=None)
+
+plt.xlabel('Temperature')
+plt.ylabel('% backbones with\n(scRMSD < 2Å)')
+
+
+# Force x-axis tick labels to show
+plt.gca().set_xticks(averages['temperature'].unique())  # Set ticks at each temperature value
+plt.gca().set_xticklabels(averages['temperature'].unique())  # Force labels to show
+
+# Remove top and right spines
+plt.gca().spines['top'].set_visible(False)
+plt.gca().spines['right'].set_visible(False)
+
+plt.tight_layout()
+plt.title('BackboneRef sample designability\nby temperature')
+

analysis/figure2b+d.py

@@ -0,0 +1,60 @@
+import pandas as pd
+import seaborn as sns
+import matplotlib.pyplot as plt
+from scipy import stats
+
+sns.set(font_scale=1.7)
+sns.set_style("whitegrid")
+plt.rcParams['axes.grid'] = False
+plt.rc('axes',edgecolor='black')
+
+plt.rc("text", usetex=False)
+plt.rc(
+    "text.latex",
+    preamble=r"\usepackage{newpxtext}\usepackage{newpxmath}\usepackage{commath}\usepackage{mathtools}",
+)
+plt.rcParams['font.family'] = 'sans-serif'
+plt.rcParams['font.sans-serif'] = ['DejaVu Sans']
+plt.rc("savefig", dpi=500)
+plt.rc("legend", loc="best", fontsize="medium", fancybox=True, framealpha=0.5)
+plt.rc("lines", linewidth=2.5, markersize=10, markeredgewidth=2.5)
+plt.rc("axes", titlepad=10)
+colors = ["#BBBBBB", "#33BBEE", "#EE3377", "#009988", "#CC3311", "#0077BB"]
+colors = list(reversed(colors))
+sns.set_palette(sns.color_palette(colors))
+# set mpl palette
+plt.rcParams["axes.prop_cycle"] = plt.cycler(color=colors)
+
+# this should be on the zenodo
+df = pd.read_parquet('backbone_novelty_quality_statistics.parquet')
+
+# figure 2d
+sns.histplot(df,
+             x='length',
+             y='max_search_tm')
+plt.ylabel('TM-score (AFDB/UniProt)')
+plt.xlabel('Backbone length (AA)')
+plt.xlim(40, 512)
+_ = plt.xticks([100, 200, 300, 400, 500, ])
+plt.ylim(0, 1.0)
+plt.title('Max. TM-score of BR structures')
+# make correlation txt
+corr_df = df[['length', 'max_search_tm']]
+corr_df.dropna(inplace=True)
+res = stats.pearsonr(corr_df['length'], corr_df['max_search_tm'])
+r = res.correlation.item()
+_ = plt.text(350, 0.85, f'R = {r:.2f}')
+
+
+# figure 2b
+plt.figure()
+sns.ecdfplot(df['avg_scrmsd'], stat='proportion', 
+             complementary=False, linewidth=2, color='#0D96C9')
+plt.xlabel('Average scRMSD')
+plt.ylabel('Percentile')
+plt.title('Designability of BR backbones')
+plt.axvline(x=2, linestyle='--', color='black',zorder=-1, alpha=0.5)
+
+
+
+

analysis/figure2c.py

@@ -0,0 +1,129 @@
+import pathlib
+import pandas as pd
+import joblib
+import numpy as np
+import seaborn as sns
+import matplotlib.pyplot as plt
+
+sns.set(font_scale=1.7)
+sns.set_style("whitegrid")
+plt.rcParams["axes.grid"] = False
+plt.rc("axes", edgecolor="black")
+plt.rcParams['font.family'] = 'sans-serif'
+plt.rcParams['font.sans-serif'] = ['DejaVu Sans']
+plt.rc("savefig", dpi=500)
+plt.rc("legend", loc="best", fontsize="medium", fancybox=True, framealpha=0.5)
+plt.rc("lines", linewidth=2.5, markersize=10, markeredgewidth=2.5)
+plt.rc("axes", titlepad=10)
+
+colors = ["#BBBBBB", "#33BBEE", "#EE3377", "#009988", "#CC3311", "#0077BB"]
+colors = list(reversed(colors))
+sns.set_palette(sns.color_palette(colors))
+plt.rcParams["axes.prop_cycle"] = plt.cycler(color=colors)
+
+together_cluster = pd.read_csv(
+    "pdb_plus_240k/scRMSD_best_240k_plus_pdbFirst_default__cluster.tsv",
+    sep="\t",
+    header=None,
+    names=["representative", "member"],
+)
+
+
+representatives, pdb_counts, percentages, sizes, member, effective_sz = (
+    [],
+    [],
+    [],
+    [],
+    [],
+    [],
+)
+for representative, groupby in together_cluster.groupby("representative"):
+    unique_pdb = groupby["member"][groupby["member"].str.startswith("pdb")].tolist()
+    unique_pdb = [s.split("_")[0] for s in unique_pdb]
+    unique_pdb = set(unique_pdb)
+
+    unique_non_pdb = groupby["member"][
+        ~groupby["member"].str.startswith("pdb")
+    ].tolist()
+    unique_non_pdb = set(unique_non_pdb)
+    percentage_pdb = len(unique_pdb) / len(unique_pdb.union(unique_non_pdb))
+
+    representatives.append(representative)
+    pdb_counts.append(len(unique_pdb))
+    percentages.append(percentage_pdb)
+    sizes.append(len(groupby))
+    effective_size = len(unique_pdb.union(unique_non_pdb))
+    effective_sz.append(effective_size)
+
+cluster_pdb = pd.DataFrame(
+    {
+        "representative": representatives,
+        "pdb_count": pdb_counts,
+        "percentage": percentages,
+        "size": sizes,
+        "effective_size": effective_sz,
+    }
+)
+cluster_pdb["pdb_log"] = np.log(cluster_pdb["pdb_count"] + 1)
+cluster_pdb["size_log"] = np.log(cluster_pdb["size"])
+cluster_pdb["size_syn"] = cluster_pdb["size"] - cluster_pdb["pdb_count"]
+
+cluster_pdb['is_synthetic'] = cluster_pdb['size_syn'] > 0
+cluster_pdb['syn_only'] = cluster_pdb['size_syn'] == cluster_pdb['size']
+
+
+
+fig, axs = plt.subplots(1, 1)
+cluster_pdb_sorted = cluster_pdb.sort_values('syn_only', ascending=True)
+
+g = sns.scatterplot(
+    data=cluster_pdb_sorted,
+    x="size",
+    y="size_syn",
+    alpha=1,
+    ax=axs,
+    s=20,
+    hue='syn_only',
+    palette=['#7f7f7f', '#0D96C9'],
+    legend=True  # Turn off automatic legend
+)
+
+f = 0.1
+xmin = cluster_pdb_sorted['size'].min() 
+xmin = xmin - xmin * f
+
+xmax = cluster_pdb_sorted['size'].max()
+xmax = xmax + xmin * f
+
+# Log scale for x-axis
+axs.set_xscale("log")
+axs.set_yscale('log')
+
+# Set labels and title
+axs.set_xlabel("Cluster size")
+axs.set_ylabel("# BBR members")
+axs.set_title("Cluster size by num. BBR members")
+
+#Create custom legend handles
+from matplotlib.lines import Line2D
+legend_elements = [
+    Line2D([0], [0], marker='o', color='w', markerfacecolor='#7f7f7f', 
+           label='PDB + BBRef samples', markersize=6),
+    Line2D([0], [0], marker='o', color='w', markerfacecolor='#0D96C9', 
+           label='BBRef samples only', markersize=6)
+]
+# PDB + BBRef samples" & "BBRef samples only" ?
+# also please change "PDB + synthetic samples" to "PDB + BBRef samples"
+
+# Add custom legend
+axs.legend(handles=legend_elements, frameon=False, markerscale=2.5)
+
+# Remove frame
+axs.spines['top'].set_visible(False)
+axs.spines['right'].set_visible(False)
+
+plt.tight_layout()
+
+
+
+