홈 › Papers

Deep learning framework DNN with conditional WGAN for protein solubility prediction

Protein solubility plays a critical role in improving production yield of recombinant proteins in biocatalyst and pharmaceutical field. To some extent, protein solubility can represent the function and activity of biocatalysts which are mainly composed of recombinant proteins. Highly soluble proteins are more effective in biocatalytic processes and can reduce the cost of biocatalysts. Screening proteins by experiments in vivo is time-consuming and expensive. In literature, large amounts of machine learning models have been investigated, whereas parameters of those models are underdetermined with insufficient data of protein solubility. A data augmentation algorithm that can enlarge the dataset of protein solubility and improve the performance of prediction model is highly desired, which can alleviate the common issue of insufficient data in biotechnology applications for developing machine learning models. We first implemented a novel approach that a data augmentation algorithm, conditional WGAN was used to improve prediction performance of DNN for protein solubility from protein sequence by generating artificial data. After adding mimic data produced from conditional WGAN, the prediction performance represented by $R^{2}$ was improved compared with the $R^{2}$ without data augmentation. After tuning the hyperparameters of two algorithms and organizing the dataset, we achieved a $R^{2}$ value of $45.04\%$, which enhanced $R^{2}$ about $10\%$ compared with the previous study using the same dataset. Data augmentation opens the door to applications of machine learning models on biological data, as machine learning models always fail to be well trained by small datasets.

Code (0)

등록된 구현이 없습니다.

Tasks

Similar Papers 제목 키워드 기반