""" Step 4: Data Visualization Generate 5 publication-quality charts for the GLP-1/GIPR portfolio strategy review. """ import matplotlib matplotlib.use('Agg') import matplotlib.pyplot as plt import matplotlib.patches as mpatches import matplotlib.ticker as ticker import numpy as np import pandas as pd from pathlib import Path import warnings warnings.filterwarnings('ignore') # ─── Paths ──────────────────────────────────────────────────────────────────── BASE = Path('/app/sandbox/session_20260403_103155_c18ae3efc2c1') RESULTS = BASE / 'results' FIGURES = BASE / 'figures' FIGURES.mkdir(parents=True, exist_ok=True) # ─── Global style ───────────────────────────────────────────────────────────── plt.rcParams.update({ 'font.family': 'sans-serif', 'font.sans-serif': ['DejaVu Sans'], 'font.size': 10, 'axes.titlesize': 13, 'axes.titleweight': 'bold', 'axes.labelsize': 11, 'axes.spines.top': False, 'axes.spines.right': False, 'xtick.labelsize': 9, 'ytick.labelsize': 9, 'legend.fontsize': 9, 'figure.dpi': 150, 'savefig.dpi': 300, 'savefig.bbox': 'tight', }) PALETTE = { 'tirzepatide': '#1f77b4', 'semaglutide': '#ff7f0e', 'retatrutide': '#2ca02c', 'CT-388': '#d62728', 'survodutide': '#9467bd', 'orforglipron': '#8c564b', 'pemvidutide': '#e377c2', 'amycretin': '#7f7f7f', } # ─── Mechanism colours (for bubble chart) ───────────────────────────────────── MECH_PALETTE = { 'mono': '#1f77b4', 'dual': '#ff7f0e', 'triple': '#2ca02c', } print("=" * 60) print("Step 4: Generating Visualizations") print("=" * 60) # ══════════════════════════════════════════════════════════════════════════════ # Chart 1 — Pipeline Bubble Chart # ══════════════════════════════════════════════════════════════════════════════ print("\n[1/5] Pipeline Bubble Chart ...") df_pipe = pd.read_csv(RESULTS / 'pipeline_data.csv') print(f" Loaded pipeline_data.csv: {df_pipe.shape[0]} rows") # Normalise phase to a numeric rank for plotting PHASE_ORDER = { 'PHASE1': 1, 'PHASE2': 2, 'PHASE3': 3, 'PHASE4 (Approved - US, EU)': 4, } def phase_rank(phase_str): for key, val in sorted(PHASE_ORDER.items(), key=lambda x: -len(x[0])): if key in phase_str: return val return 0 df_pipe['phase_num'] = df_pipe['phase'].apply(phase_rank) df_pipe['phase_label'] = df_pipe['phase_num'].map({ 1: 'Phase 1', 2: 'Phase 2', 3: 'Phase 3', 4: 'Approved' }) # Mechanism category def mech_cat(m): m = m.lower() if 'triple' in m: return 'triple' if 'dual' in m: return 'dual' return 'mono' df_pipe['mech_cat'] = df_pipe['mechanism'].apply(mech_cat) # Route marker def route_marker(r): r = r.lower() if 'oral' in r and 'inject' in r: return 'D' # diamond — both if 'oral' in r: return 's' # square return 'o' # circle (injectable) df_pipe['marker'] = df_pipe['route'].apply(route_marker) # Bubble size ~ total_trials (scale for visibility) sizes = (df_pipe['total_trials'].clip(lower=1) / df_pipe['total_trials'].max()) * 1200 + 200 fig, ax = plt.subplots(figsize=(10, 5.5)) for _, row in df_pipe.iterrows(): color = MECH_PALETTE[row['mech_cat']] ax.scatter( row['phase_num'], 0.5, s=sizes[row.name], c=color, marker=row['marker'], alpha=0.80, edgecolors='white', linewidths=1.2, zorder=3 ) ax.text( row['phase_num'], 0.5 + 0.08, row['drug_name'], ha='center', va='bottom', fontsize=8.5, fontweight='bold' ) ax.set_xlim(0.5, 4.5) ax.set_ylim(0, 1) ax.set_xticks([1, 2, 3, 4]) ax.set_xticklabels(['Phase 1', 'Phase 2', 'Phase 3', 'Approved'], fontsize=10) ax.set_yticks([]) ax.spines['left'].set_visible(False) ax.set_xlabel('Development Phase', fontsize=11) ax.set_title('GLP-1/Incretin Agonist Competitive Pipeline', pad=14) # Legend: mechanism colours legend_mech = [ mpatches.Patch(color=MECH_PALETTE['mono'], label='Mono agonist (GLP-1R)'), mpatches.Patch(color=MECH_PALETTE['dual'], label='Dual agonist (GLP-1R/GIPR or GCGR)'), mpatches.Patch(color=MECH_PALETTE['triple'], label='Triple agonist (GLP-1R/GIPR/GCGR)'), ] # Legend: route markers (dummy scatter for legend handles) h_inj = plt.scatter([], [], s=80, c='grey', marker='o', label='Injectable') h_oral = plt.scatter([], [], s=80, c='grey', marker='s', label='Oral') h_both = plt.scatter([], [], s=80, c='grey', marker='D', label='Oral & Injectable') leg1 = ax.legend(handles=legend_mech, loc='upper left', title='Mechanism', framealpha=0.9) ax.add_artist(leg1) ax.legend(handles=[h_inj, h_oral, h_both], loc='upper right', title='Route', framealpha=0.9) # Bubble size note ax.annotate('Bubble size proportional to number of registered trials', xy=(0.5, 0.02), xycoords='axes fraction', fontsize=7.5, color='grey', ha='left') fig.tight_layout() out1 = FIGURES / 'pipeline_bubble_chart.png' fig.savefig(out1, dpi=300, bbox_inches='tight') plt.close(fig) print(f" Saved: {out1}") # ══════════════════════════════════════════════════════════════════════════════ # Chart 2 — Efficacy Bar Chart # ══════════════════════════════════════════════════════════════════════════════ print("\n[2/5] Efficacy Bar Chart ...") df_eff = pd.read_csv(RESULTS / 'clinical_efficacy_summary_std_nonT2D.csv') print(f" Loaded clinical_efficacy_summary_std_nonT2D.csv: {df_eff.shape[0]} rows") print(f" Columns: {list(df_eff.columns)}") # Sort descending by efficacy df_eff_sorted = df_eff.sort_values('Placebo_Adjusted_Weight_Loss_Pct', ascending=False).reset_index(drop=True) colors = [PALETTE.get(d.lower().replace('-', '-'), '#777') for d in df_eff_sorted['Drug']] # Manual colour lookup bar_colors = [] for drug in df_eff_sorted['Drug']: bar_colors.append(PALETTE.get(drug, '#777777')) fig, ax = plt.subplots(figsize=(8, 5.5)) bars = ax.bar( df_eff_sorted['Drug'], df_eff_sorted['Placebo_Adjusted_Weight_Loss_Pct'], color=bar_colors, edgecolor='white', linewidth=1, width=0.55, zorder=3 ) ax.set_ylim(0, df_eff_sorted['Placebo_Adjusted_Weight_Loss_Pct'].max() * 1.35) ax.set_ylabel('Placebo-Adjusted Mean Body\nWeight Loss (%)', fontsize=10) ax.set_title('Head-to-Head Efficacy Comparison\n(Non-T2D Obesity Trials, Standard Design)', pad=12) ax.yaxis.grid(True, linestyle='--', alpha=0.4, zorder=0) ax.set_axisbelow(True) # Value labels on bars for bar, val in zip(bars, df_eff_sorted['Placebo_Adjusted_Weight_Loss_Pct']): ax.text( bar.get_x() + bar.get_width() / 2, bar.get_height() + 0.3, f'{val:.1f}%', ha='center', va='bottom', fontsize=9.5, fontweight='bold' ) # Duration labels below drug name (x-tick) duration_map = dict(zip(df_eff_sorted['Drug'], df_eff_sorted['Duration_Weeks'])) new_labels = [] for drug in df_eff_sorted['Drug']: wks = int(duration_map[drug]) new_labels.append(f'{drug}\n({wks} wks)') ax.set_xticklabels(new_labels, fontsize=9.5) # ── SCIENTIFIC CAVEAT — CT-388 ──────────────────────────────────────────────── caveat_text = ( "⚠ IMPORTANT CAVEAT: CT-388 data represent a 24-week\n" " interim analysis only. All other drugs reflect 48–72\n" " week endpoints. Direct comparisons should be interpreted\n" " with caution — efficacy at full trial duration may differ." ) ax.text( 0.98, 0.97, caveat_text, transform=ax.transAxes, fontsize=8, color='#8B0000', ha='right', va='top', bbox=dict(boxstyle='round,pad=0.4', facecolor='#fff8f8', edgecolor='#8B0000', linewidth=1.2, alpha=0.95), fontweight='normal', linespacing=1.45 ) fig.tight_layout() out2 = FIGURES / 'efficacy_bar_chart.png' fig.savefig(out2, dpi=300, bbox_inches='tight') plt.close(fig) print(f" Saved: {out2}") # ══════════════════════════════════════════════════════════════════════════════ # Chart 3 — FAERS Time-Series # ══════════════════════════════════════════════════════════════════════════════ print("\n[3/5] FAERS Time-Series ...") df_faers = pd.read_csv(RESULTS / 'faers_time_series.csv') print(f" Loaded faers_time_series.csv: {df_faers.shape[0]} rows") print(f" Event types: {sorted(df_faers['event_type'].unique())}") print(f" Drugs: {sorted(df_faers['drug'].unique())}") # Create a proper date from quarter string (e.g. "2020Q1" -> 2020-01-01) def quarter_to_date(q): yr, qt = q[:4], int(q[5]) month = (qt - 1) * 3 + 1 return pd.Timestamp(year=int(yr), month=month, day=1) df_faers['date'] = df_faers['quarter'].apply(quarter_to_date) # Aggregate GI events (nausea + vomiting + diarrhea + abdominal_pain) and pancreatitis GI_EVENTS = {'nausea', 'vomiting', 'diarrhea', 'abdominal_pain'} # Aggregate per drug × date gi_agg = ( df_faers[df_faers['event_type'].isin(GI_EVENTS)] .groupby(['date', 'drug'], as_index=False) .agg(event_count=('event_count', 'sum'), total_drug_reports=('total_drug_reports_in_quarter', 'first')) ) gi_agg['rate'] = gi_agg['event_count'] / gi_agg['total_drug_reports'] * 100 panc_agg = ( df_faers[df_faers['event_type'] == 'pancreatitis'] .groupby(['date', 'drug'], as_index=False) .agg(event_count=('event_count', 'sum'), total_drug_reports=('total_drug_reports_in_quarter', 'first')) ) panc_agg['rate'] = panc_agg['event_count'] / panc_agg['total_drug_reports'] * 100 DRUGS_FAERS = ['semaglutide', 'tirzepatide'] LINE_STYLES_FAERS = {'semaglutide': '-', 'tirzepatide': '--'} MARKERS_FAERS = {'semaglutide': 'o', 'tirzepatide': 's'} fig, (ax_gi, ax_panc) = plt.subplots(1, 2, figsize=(13, 5), sharey=False) fig.suptitle('FAERS Adverse Event Reporting Rates: Semaglutide vs. Tirzepatide (2020–2026)', fontsize=12, fontweight='bold', y=1.01) for drug in DRUGS_FAERS: sub = gi_agg[gi_agg['drug'] == drug].sort_values('date') if sub.empty: continue ax_gi.plot( sub['date'], sub['rate'], color=PALETTE[drug], linestyle=LINE_STYLES_FAERS[drug], marker=MARKERS_FAERS[drug], markersize=4, linewidth=1.8, label=drug.capitalize() ) ax_gi.set_title('GI Adverse Events\n(Nausea, Vomiting, Diarrhea, Abdominal Pain)', fontsize=10) ax_gi.set_ylabel('Normalized Reporting Rate (%)', fontsize=10) ax_gi.set_xlabel('Quarter', fontsize=9) ax_gi.legend(framealpha=0.85) ax_gi.yaxis.grid(True, linestyle='--', alpha=0.35) ax_gi.xaxis.set_major_formatter(plt.matplotlib.dates.DateFormatter('%Y')) ax_gi.xaxis.set_major_locator(plt.matplotlib.dates.YearLocator()) for drug in DRUGS_FAERS: sub = panc_agg[panc_agg['drug'] == drug].sort_values('date') if sub.empty: continue ax_panc.plot( sub['date'], sub['rate'], color=PALETTE[drug], linestyle=LINE_STYLES_FAERS[drug], marker=MARKERS_FAERS[drug], markersize=4, linewidth=1.8, label=drug.capitalize() ) ax_panc.set_title('Pancreatitis Reporting Rate', fontsize=10) ax_panc.set_ylabel('Normalized Reporting Rate (%)', fontsize=10) ax_panc.set_xlabel('Quarter', fontsize=9) ax_panc.legend(framealpha=0.85) ax_panc.yaxis.grid(True, linestyle='--', alpha=0.35) ax_panc.xaxis.set_major_formatter(plt.matplotlib.dates.DateFormatter('%Y')) ax_panc.xaxis.set_major_locator(plt.matplotlib.dates.YearLocator()) # Note tirzepatide FDA approval ax_gi.axvline(pd.Timestamp('2022-05-01'), color='grey', linestyle=':', linewidth=1, alpha=0.7) ax_gi.text(pd.Timestamp('2022-05-01'), ax_gi.get_ylim()[1]*0.98, 'Tirzepatide\nFDA 2022', fontsize=7, color='grey', ha='left', va='top') ax_panc.axvline(pd.Timestamp('2022-05-01'), color='grey', linestyle=':', linewidth=1, alpha=0.7) fig.tight_layout() out3 = FIGURES / 'faers_time_series.png' fig.savefig(out3, dpi=300, bbox_inches='tight') plt.close(fig) print(f" Saved: {out3}") # ══════════════════════════════════════════════════════════════════════════════ # Chart 4 — Pharmacology Scatter Plot # ══════════════════════════════════════════════════════════════════════════════ print("\n[4/5] Pharmacology Scatter Plot ...") df_pharm = pd.read_csv(RESULTS / 'pharmacology_data.csv') print(f" Loaded pharmacology_data.csv: {df_pharm.shape[0]} rows") # Keep only rows with numeric values; pivot to wide format df_valid = df_pharm[df_pharm['value_nM'].notna() & df_pharm['value_nM'].astype(str).str.strip().ne('')].copy() df_valid['value_nM'] = pd.to_numeric(df_valid['value_nM'], errors='coerce') df_valid = df_valid.dropna(subset=['value_nM']) # Keep literature-sourced rows (de-duplicate: use first best entry per drug × target) df_lit = df_valid[df_valid['source'].str.contains('Literature', na=False)].copy() df_lit = df_lit.sort_values('value_nM').drop_duplicates(subset=['drug_name', 'target'], keep='first') # Pivot to GLP-1R vs GIPR df_glp1 = df_lit[df_lit['target'] == 'GLP-1R'][['drug_name', 'value_nM', 'pchembl_value']].rename( columns={'value_nM': 'glp1r_ec50', 'pchembl_value': 'glp1r_pchembl'}) df_gipr = df_lit[df_lit['target'] == 'GIPR'][['drug_name', 'value_nM', 'pchembl_value']].rename( columns={'value_nM': 'gipr_ec50', 'pchembl_value': 'gipr_pchembl'}) df_wide = pd.merge(df_glp1, df_gipr, on='drug_name', how='left') # For mono-GLP-1R drugs, assign a placeholder GIPR value far right to show selectivity # Use log10(EC50) for axes df_wide['log_glp1r'] = np.log10(df_wide['glp1r_ec50']) df_wide['log_gipr'] = np.where( df_wide['gipr_ec50'].notna(), np.log10(df_wide['gipr_ec50']), np.nan ) # Only plot drugs with GLP-1R data df_plot = df_wide.dropna(subset=['log_glp1r']).copy() print(f" Drugs with GLP-1R EC50: {list(df_plot['drug_name'])}") fig, ax = plt.subplots(figsize=(8, 6)) for _, row in df_plot.iterrows(): drug = row['drug_name'] color = PALETTE.get(drug, '#666666') has_gipr = pd.notna(row['log_gipr']) if has_gipr: ax.scatter(row['log_glp1r'], row['log_gipr'], color=color, s=140, zorder=4, edgecolors='white', linewidths=0.8, label=drug) ax.annotate(drug, (row['log_glp1r'], row['log_gipr']), textcoords='offset points', xytext=(6, 4), fontsize=8.5, fontweight='bold', color=color) else: # Show on right axis margin as "GIPR inactive" ax.scatter(row['log_glp1r'], ax.get_ylim()[0] if ax.get_ylim()[0] < -2 else -2.5, color=color, s=140, marker='v', zorder=4, edgecolors='white', linewidths=0.8, label=f'{drug} (GIPR inactive)') ax.annotate(f'{drug}\n(GIPR n/a)', (row['log_glp1r'], -2.5), textcoords='offset points', xytext=(5, -12), fontsize=7.5, color=color) # Diagonal line of equal potency (balanced agonism) xmin = df_plot['log_glp1r'].min() - 0.5 xmax = df_plot['log_glp1r'].max() + 0.5 ax.plot([xmin, xmax], [xmin, xmax], 'k--', linewidth=0.9, alpha=0.4, label='Equal potency (1:1)') ax.set_xlabel('log\u2081\u2080(GLP-1R EC\u2085\u2080, nM)\n← More potent', fontsize=10) ax.set_ylabel('log\u2081\u2080(GIPR EC\u2085\u2080, nM)\n← More potent', fontsize=10) ax.set_title('Receptor Selectivity: GLP-1R vs. GIPR Potency\n(Lower = More Potent)', pad=12) ax.legend(fontsize=8, loc='upper left', framealpha=0.85, ncol=1) ax.yaxis.grid(True, linestyle='--', alpha=0.35) ax.xaxis.grid(True, linestyle='--', alpha=0.35) # Annotate CT-388 balanced region ct388_row = df_plot[df_plot['drug_name'] == 'CT-388'] if not ct388_row.empty: ax.annotate( 'CT-388:\nBalanced GLP-1R/GIPR', xy=(ct388_row['log_glp1r'].values[0], ct388_row['log_gipr'].values[0]), xytext=(-55, 25), textcoords='offset points', arrowprops=dict(arrowstyle='->', color='#d62728', lw=1.2), fontsize=8, color='#d62728', fontweight='bold' ) fig.tight_layout() out4 = FIGURES / 'pharmacology_scatter.png' fig.savefig(out4, dpi=300, bbox_inches='tight') plt.close(fig) print(f" Saved: {out4}") # ══════════════════════════════════════════════════════════════════════════════ # Chart 5 — Patent Gantt Chart # ══════════════════════════════════════════════════════════════════════════════ print("\n[5/5] Patent Gantt Chart ...") # Curated patent expiry data (well-known from public sources) # Sources: FDA Orange Book, USPTO, company filings, published analyses patent_data = [ # drug, type, territory, start_year, expiry_year, notes ('Semaglutide', 'Compound patent', 'US', 2012, 2031, 'Core composition-of-matter'), ('Semaglutide', 'Compound patent', 'EU', 2012, 2031, 'Core composition-of-matter'), ('Semaglutide', 'Formulation/method patent', 'US', 2017, 2034, 'Oral formulation (Rybelsus)'), ('Semaglutide', 'Formulation/method patent', 'EU', 2017, 2033, 'Injectable device/formulation'), ('Tirzepatide', 'Compound patent', 'US', 2015, 2036, 'Core composition-of-matter'), ('Tirzepatide', 'Compound patent', 'EU', 2015, 2036, 'Core composition-of-matter'), ('Tirzepatide', 'Method-of-use patent', 'US', 2020, 2040, 'Obesity/CVD indication claims'), ('Retatrutide', 'Compound patent', 'US', 2018, 2038, 'Estimated from priority date'), ('Retatrutide', 'Compound patent', 'EU', 2018, 2038, 'Estimated from priority date'), ('CT-388', 'Compound patent', 'US', 2020, 2040, 'Estimated from priority filing'), ('CT-388', 'Compound patent', 'EU', 2020, 2040, 'Estimated from priority filing'), ] df_patent = pd.DataFrame(patent_data, columns=['Drug', 'Patent_Type', 'Territory', 'Start_Year', 'Expiry_Year', 'Notes']) # Sort by expiry df_patent_sorted = df_patent.sort_values(['Drug', 'Territory', 'Start_Year']).reset_index(drop=True) # Build Gantt chart DRUG_ORDER = ['Semaglutide', 'Tirzepatide', 'Retatrutide', 'CT-388'] TYPE_COLORS = { 'Compound patent': '#1f77b4', 'Formulation/method patent': '#ff7f0e', 'Method-of-use patent': '#2ca02c', } TERR_HATCH = {'US': '', 'EU': '///'} fig, ax = plt.subplots(figsize=(12, 6)) y = 0 yticks = [] ytick_labels = [] bar_height = 0.35 for drug in DRUG_ORDER: sub = df_patent_sorted[df_patent_sorted['Drug'] == drug] for _, row in sub.iterrows(): color = TYPE_COLORS.get(row['Patent_Type'], '#888') hatch = TERR_HATCH.get(row['Territory'], '') bar = ax.barh( y, row['Expiry_Year'] - row['Start_Year'], left=row['Start_Year'], height=bar_height, color=color, hatch=hatch, edgecolor='white', linewidth=0.6, alpha=0.85 ) # Label: expiry year at end of bar ax.text(row['Expiry_Year'] + 0.2, y, str(row['Expiry_Year']), va='center', fontsize=7.5, color='#333') # Territory label ax.text(row['Start_Year'] + 0.4, y, row['Territory'], va='center', fontsize=7, color='white', fontweight='bold') yticks.append(y) ytick_labels.append(f"{drug}\n{row['Patent_Type']}") y -= 0.55 y -= 0.2 # gap between drugs ax.set_xlabel('Year', fontsize=11) ax.set_title('Competitive Patent Expiry Timelines\n(Key GLP-1/Incretin Programs)', pad=12) ax.set_yticks(yticks) ax.set_yticklabels(ytick_labels, fontsize=8) ax.set_xlim(2010, 2045) ax.xaxis.set_major_locator(ticker.MultipleLocator(5)) ax.xaxis.grid(True, linestyle='--', alpha=0.35) # Today line ax.axvline(2026, color='red', linewidth=1.2, linestyle='--', alpha=0.7) ax.text(2026.2, ax.get_ylim()[1] * 0.99 if ax.get_ylim()[1] < 0 else yticks[0] + 0.3, 'Today\n(2026)', color='red', fontsize=8, va='top') # Patent cliff annotation ax.axvspan(2031, 2033, alpha=0.07, color='orange') ax.text(2031.5, min(yticks) - 0.5, 'Semaglutide\npatent cliff', fontsize=7.5, color='darkorange', ha='center', va='top') # Legend for colors legend_handles = [ mpatches.Patch(color=c, label=t) for t, c in TYPE_COLORS.items() ] legend_handles += [ mpatches.Patch(facecolor='white', edgecolor='black', hatch='', label='US territory'), mpatches.Patch(facecolor='white', edgecolor='black', hatch='///', label='EU territory'), ] ax.legend(handles=legend_handles, loc='lower right', fontsize=8, framealpha=0.9, title='Patent Type / Territory', title_fontsize=8) ax.text(0.01, -0.12, 'Note: Estimated expiry years based on priority filing dates and standard 20-year patent term.\n' 'Actual dates may vary due to patent term extensions (PTE), SPCs, and litigation outcomes.', transform=ax.transAxes, fontsize=7, color='grey', ha='left') fig.tight_layout() out5 = FIGURES / 'patent_gantt_chart.png' fig.savefig(out5, dpi=300, bbox_inches='tight') plt.close(fig) print(f" Saved: {out5}") # ─── Final check ────────────────────────────────────────────────────────────── print("\n" + "=" * 60) print("Visualization generation complete.") generated = [out1, out2, out3, out4, out5] for p in generated: size_kb = p.stat().st_size / 1024 if p.exists() else 0 status = "OK" if p.exists() else "MISSING" print(f" [{status}] {p.name} ({size_kb:.0f} KB)") print("=" * 60)