The copy number of genes in chromosomes can be modified by chromosomal integration to construct efficient microbial cell factories but the resulting genetic systems are prone to failure or instability from triggering homologous recombination in repetitive DNA sequences. Finding the optimal copy number of each gene in a pathway is also time and labor intensive. To overcome these challenges, we applied a multiple nonrepetitive coding sequence calculator that generates sets of coding DNA sequence (CDS) variants. A machine learning method was developed to calculate the optimal copy number combination of genes in a pathway. We obtained an engineered Yarrowia lipolytica strain for eicosapentaenoic acid biosynthesis in 6 months, producing the highest titer of 27.5 g l⁻¹ in a 50-liter bioreactor. Moreover, the lycopene production in Escherichia coli was also greatly improved. Importantly, all engineered strains of Y. lipolytica, E. coli and Saccharomyces cerevisiae constructed with nonrepetitive CDSs maintained genetic stability.

染色体中基因的拷贝数可以通过染色体整合进行修改，以构建有效的微生物细胞工厂，但由此产生的遗传系统很容易因触发重复DNA序列中的同源重组而失败或不稳定。寻找通路中每个基因的最佳拷贝数也是时间和劳动力密集型的。为了克服这些挑战，我们应用了多个非重复编码序列计算器来生成编码 DNA 序列 (CDS) 变体集。开发了一种机器学习方法来计算通路中基因的最佳拷贝数组合。我们在 6 个月内获得了用于二十碳五烯酸生物合成的工程化解脂耶氏酵母菌株，在 50 升生物反应器中产生了 27.5 gl ^-1的最高滴度。而且，大肠杆菌中番茄红素的产量也大大提高。重要的是，所有用非重复CDS构建的解脂耶氏酵母、大肠杆菌和酿酒酵母工程菌株都保持了遗传稳定性。