Skip to content

Output Format

Directory Structure

output/<patient_id>/
├── oncokb_output.tsv
├── pubmed_output.tsv
├── txgnn_output.tsv
├── clinicaltrials_output.tsv
├── top20_drugs.png
├── txgnn_graph.png
├── final_report.xlsx
└── final_report.pdf

Excel Report Sheets

The final_report.xlsx contains seven sheets:

Sheet Description
Summary Cancer type, variant count, and top 20 ranked drugs
Performance Runtime, API calls, and memory usage per module
OncoKB Full OncoKB annotation output
PubMed Gene-drug mention counts from PubMed
TxGNN Graph-based drug candidates with category labels
ClinicalTrials Raw clinical trial records
Merged_Drugs Final merged and ranked drug table

Merged Drug Table Columns

The Merged_Drugs sheet contains the final ranked output:

Column Description
drug Drug name (uppercase)
combined_score Percentile-normalized combined score
tx_pct TxGNN percentile
pm_pct PubMed percentile
ok_pct OncoKB percentile
support_count Number of evidence layers (0--3)
clinical_flag Has matched clinical trials (0/1)
source Evidence source labels (e.g., TxGNN\|PubMed)
txgnn_score Raw TxGNN score
mention_count Raw PubMed mention count
oncokb_score Raw OncoKB score
n_clinical_trials Number of matched trials
top_nct_id NCT ID of highest-phase trial
top_phase Highest trial phase

PDF Report

The PDF report includes:

  1. Summary page -- Cancer type, variant count, top drugs
  2. Top 20 drugs bar plot -- Ranked by combined score
  3. TxGNN network graph -- Cancer type, mutated genes, and recommended drugs
  4. Clinical trials table -- Top 30 drugs with trial metadata

Visualizations

Top 20 Drugs (top20_drugs.png)

Horizontal bar chart of the 20 highest-scoring drugs by combined score.

TxGNN Network Graph (txgnn_graph.png)

Network visualization with:

  • Red nodes: Cancer type
  • Blue nodes: Mutated genes
  • Green nodes: Recommended drugs
  • Edges: Gene-drug associations from the knowledge graph