. 2015 Mar 25;10(3):e0119294. doi: 10.1371/journal.pone.0119294

Table 1. Algorithm 1.

The pseudocode procedure for the parallel communications of the LSGPA.
*for* Slave i = 1 to i = M *in Parrallel*
→.Let Slave(i) *MPI_Rev* dataset(i) from Master, C _i;
→.MPI_Brocast C _i to all Slaves for 1 to M;
→.Compute I(x,y) for every pair of genes (x,y);
→.Give the *threshold* of MI values λ, if I(x,y)<λ then I(x,y) = 0,
else I(x,y) = 1 and edge (x,y) will be selected into C _i (Candidate_set(i));
→.MPI_Rev C _i from Slaves1 to M
*Compute* MI values for every pair of edges (x,y _j) x∈C _i, y _j∈C _k,k = 1 to M, k≠i;
*Rank* all the MI values, selected top 5% value for every pair of edges into Candidate_set of C _i.
→.Give the paameter threshold θ(**)
*for* j = 1 to T _i
Solving the normal equations $H_{i}^{T} H_{i} P_{i}^{j} = H_{i}^{T} D_{i}^{j} (t)$ or using the QR factorization;
to obtain parameters $P_{i}^{j} = {(a_{i 1}^{j}, a_{i 2}^{j}, L, a_{i T_{i}}^{j}, b_{i 1}^{j}, b_{i 2}^{j}, L, b_{i H_{i}}^{j})}^{T} = {(β_{1}, β_{2}, L, β 1, β_{T_{i} + H_{i}})}^{T};$
*for* l = 1 to T _i +H _i
if β _l > θ
β _l = 1 and edge(y _k,x _l) = 1;
*else*
β _l = 1 and edge(y _k,x _l) = 0;
*Endfor*
Net(i).Matrix $(K, :) = {(β_{1}, β_{2}, L, β_{T_{i} + H_{i}})}^{T};$
*Endfor*
→MPI_Send Net(i).Matrix to Master;
*Endfor*