Research On Protein Thermostability Based On Computation Intelligence Technology

The research of protein thermostability has been vigorously studied in the field ofbiophysical and biological technology. This is mainly because mesophilic protein is easy to bedenatured at high temperatures. In the process of industrial production, the phenomenon ofprotein denaturation caused difficulty to broaden its application field. Therefore, how toimprove the thermostability of protein has been concerned in the field of molecular biology,bioengineering and chemical industries. Currently, the discrimination of protein stability can beroughly divided into two kinds: one kind is using the protein spatial structure information topredict protein stability; Another kind is using the amino acid sequence information to predictits stability. The method on the spatial structure information generally has high predict accuracybecause of its known spatial structure, but for most of the protein speaking, we know only theirsequence instead of its structure, but amino acid sequence contains all the information whichdecides protein structure, so understanding the function of proteins from sequence informationis possible. In this way, we can use amino acid sequence to detect protein thermostability. In thisresearch field, how to extract features from amino acid sequence and search an effective tool torecognize mesophilic and thermophilic proteins are what we should consider.Extract features from amino acid sequence: In this paper, a feature extraction method cameto be true which considered the fusion of amino acids models and chem-composition models.Meanwhile, the feed forward artificial neural network which was optimized by particle swarmoptimization (PSO-NN) was introduced to discriminate protein thermostsbility. Compared withthe previous research, the discrimination accuracy had been improved to some extent.Meanwhile we made use of a modified Chou’s pseudo-amino acid composition method toextract features from protein sequences. That’s because many physical and chemical propertieshave business with the thermal stability, but the traditional pseudo amino acid composition onlycontains hydrophobic features, hydrophilic features and side chains atomic weight, but themodified pseudo-amino acid composition combines amino acid composition and Z-scale, thatis to say, the latter λ ones are Z-scales. Z-scales are derived from a principal componentsanalysis of a matrix of29physicochemical variables for the20coded amino acids. That is to say, the three z values can be regarded as “principal properties” of the amino acids summarizingall29measurements.We try to use new integrated feed forward artificial neural network which was optimizedby particle swarm optimization (PSO-NN) to recognize the mesophilic and thermophilicproteins. Here, we adopted Genetic Algorithm based Selected Ensemble (GASEN) as ourintegration methods. A better accuracy was got by GASEN. So, the integrated methods wereproved to be effectual. Meanwhile we made use of a modified Chou’s pseudo-amino acidcomposition method to extract features from protein sequences and Flexible Neural Tree (FNT)was firstly used as the classifier to discriminate thermophilic and mesophilic proteins.