Algorithm 1.

BiLSTM for prediction of potential pan-cancer related genes.

Input: The PPI network G = (V, E), protein complex, subcellular localization, gene expression data, threshold

Output: The classification label for proteins;

1: Calculate PC for each protein by using Equation (1);

2: Calculate NNSL for each protein by using Equation (4);

3: Calculate PeC for each protein by using Equation (6);

4: Calculate NNC for each protein by using Equation (9);

5: Incorporate PC, NNSL, PeC and NNC by using the adaptive parameters through Equation 10 to obtain PSGN score;

6: Integrate protein feature representation enhanced by [NNC, NNSL, PC, PeC, PSGN];

7: for i (1 → n) do

8: Random select (1/n)% data as training dataset, others as test dataset;

9: Classification and fix the protein label by BiLSTM;

10: end

11: Count the frequency of the predicted essential gene (labeled 1)

12: return The genes with frequencies greater than the threshold.