A Data-driven Method for BioBrick Quality Assessment

BioBrick standard biological part (shortened as BioBrick), a widely used gene segment, faces severe quality challenges due to the crowdsourcing data collection process. The complex and expensive quality control at the sequencing level, however, has not been able to check a large number of samples effectively. To tackle this problem, this paper proposes a data-driven method for the quantitative assessment of BioBrick's quality. This method only needs the data itself without the requirement for quality labels in advance, and it generates a score for each BioBrick that explains its accuracy under sequence recognition. First, a group of one-vs-rest classifiers is constructed and compared with multiple learning strategies for sequence analysis. The accuracy score is then calculated by leveraging the divergence of these classifiers. Experimental results demonstrate that the proposed method achieves good performance on BioBricks classification and quality assessment. Furthermore, the range of the most informative subregion of BioBricks narrows down to 1-150 base positions, of which 5 base positions can achieve more than 72.9% accuracy for BioBrick classification. By formulating data-driven quality calculator, this paper highlights the application of machine learning to facilitate BioBrick quality assessment in a more generalized fashion.