همایش ملی بیوانفورماتیک ایران

صفحه اصلی / 4th international edition and 13th Iranian Conference on Bioinformatics

Designing Guide RNAs Considering Essential Genes for Genome Editing of Yarrowia lipolytica Using Deep Learning in the CRISPR System

نویسندگان :

Zahra Vahdani¹ Farshad Darvishi² Fatemeh Zare-Mirakabad³

1- Department of Computer Science, Faculty of Mathematical Sciences, Alzahra University, Tehran, Iran 2- Department of Microbiology, Faculty of Biological Sciences, Alzahra University, Tehran, Iran 3- Department of Mathematics and Computer Science, Amirkabir University of Technology, Tehran, Iran

کلمات کلیدی :

Guide RNA،Deep learning،Yeast; Genome editing،vital genes

چکیده :

The CRISPR/Cas system is used to precisely remove and add one or more genes to the genome. In this system, a protein called Cas is combined with a short RNA called sgRNA, which makes a double-strand break precisely at the desired location in the genome. The designed sgRNA should be designed in a way that not only accurately targets the desired location without off-target effects but also avoids affecting vital genes. The yeast Yarrowia lipolytica can produce valuable natural and recombinant compounds with commercial, industrial, and therapeutic significance. Given the importance and application of this yeast, designing appropriate sgRNAs can yield optimal efficiency for genome editing and the production of economically valuable products. The cutting score (CS) and fitness score (FS), which indicate changes in gene activity following sgRNA deletion, were obtained in the laboratory for each sgRNA sequence in the Cas12a protein by Ramesh et al. In this study, using these values and deep learning based on a convolutional neural network (CNN), unsupervised learning was first performed with a convolutional autoencoder (CAE) to extract sgRNA features in the Y.lipolytica genome. Then, supervised learning by the CNN yielded the FS value for each sgRNA in the Cas12a dataset, resulting in Spearman values of 0.70% and Pearson values of 0.72%. The FS results for each sgRNA sequence were fed into a neural network to predict the CS, which indicates sgRNA effectiveness. Finally, the model’s predictions achieved Spearman values of 0.96% and Pearson values of 0.95% for predicting the sgRNA with the highest efficacy, outperforming existing algorithms for Y.lipolytica.