A Two-Stage Exon Recognition Model Based on Synergetic Neural Network

Algorithm 2

Precise exon regions based on adaptive smoothing window.

(1) Let 𝑊 is a given gene sequence, 𝑊_start and 𝑊_end are the beginning and the end of

  the sequence respectively;

(2) Using 𝑍_Curve mapping converted gene sequence [𝑊_start,𝑊_end] to the

  corresponding numeric sequence;

(3) Using fast Fourier transform to get spectral values;

(4) Calculating gene sequence order parameter:

    𝜉₁=𝑅₁𝑅₁+𝑅₂,   𝜉₂=𝑅₂𝑅₁+𝑅₂;

(5) Order parameter evolution according to formula (14);

(6) If 𝜉₁>𝑇₀, 𝜉₂>𝑇₀ and 𝑊_start+𝛾<𝑊_end−𝛾, Then 𝑊_start=𝑊_start+𝛾,

   𝑊_end=𝑊_end−𝛾,

   Repeat step 2 to step 6;

(7) Output the final interval [𝑊_start,𝑊_end].