Algorithm A2 Recommended Model Evaluation Algorithm

1:  Begin  2:  Initialize precision_list, recall_list, f1_list, hit_list as empty lists  3:  tested ← 0  4:  for each combination in combo_data do  5:    if size of combination < 3 then  6:      Skip the combination  7:    end if  8:    Randomly select 2 drugs from combination as known_drugs  9:    true_drugs ← remaining drugs from combination 10:    recommended ← GreedyDrugRecommendation(known_drugs, combo_freq_dict, risk_dict, drug_gene_map, top_k) 11:   recommended_set ← set of recommended drugs 12:   true_positives ← recommended_set ∩ true_drugs 13:   precision ← |true_positives|/top_k 14: recall ← |true_positives|/|true_drugs| 15:   if precision + recall > 0 then 16:      f1 ← 2 × precision × recall/(precision + recall) 17:   else 18:      f1 ← 0 19:   end if 20:   hit ← 1 if true_positives not empty else 0 21:   Append precision, recall, f1, hit to respective lists 22:   tested ← tested + 1 23:   if tested ≥ sample_size then 24:      Break 25:   end if 26:  end for 27: Calculate mean(precision_list), mean(recall_list), mean(f1_list), mean(hit_list) 28: Output EvaluationMetrics as (mean precision, mean recall, mean f1, mean hit) 29: End