Diagnose Mutations Causes Î’-Thalassemia: Biomining Method Using an Optimal Neural Learning Algorithm

Abstract

The problems in genome and proteome classification of mutations causing a thalassemia are synthesis, e.g. which thalassemia's database will choose? and then the technique that used in biomining to classify mutations causing thalassemia who can say is effective/optimal. This paper proposed genomics classification for β-thalassemia’s mutations in ITHALNET-IthaGenes database [1] (which is a modern and more comprehensive comparing to other  thalassemia databases about  63% of thalassemia’s  mutations) using data biomining method based on multiple  neural network learning algorithms (Conjugate Gradient Descent, quick propagation, online backpropagation BP and batch BP algorithm).  The experimental results based on architecture of BP [457-228-1] with (1000) iteration shows conjugate gradient descent is optimal biomining technique comparing to other techniques  of diagnosis mutation of B-thalassemia, which shows in training stage with error improvement= 5.20E-08 and testing stage Correlate= 0.999601 &  R-Squared= 0.9992, in quick propagation gives error improvement= 5.20E-08, Correlate= 0.997086 &  R-Squared= 0.994173,  in Batch BP reveals error improvement = 0.257249, Correlate= 0.975762 &  R-Squared= 0.931719,  finally the online propagation  error improvement= 0.000013 and testing stage Correlate= 0.975277 &  R-Squared= 0.900057).