CRISPR-Cas9 guide RNA efficiency prediction with efficiently tuned models in Amazon SageMaker

This article discusses using efficiently tuned models in Amazon SageMaker for predicting the efficiency of CRISPR-Cas9 guide RNA (gRNA) sequences, which is important for gene editing technologies.

Specifically, the article covers:

• Overview of the CRISPR-Cas9 mechanism and importance of predicting gRNA efficiency
• Using a pre-trained genomic language model (DNABERT) and encoding gRNA sequences
• Applying Parameter-Efficient Fine-Tuning (PEFT) methods, specifically LoRA, to reduce computational requirements
• Dataset preparation and model architecture for gRNA efficiency prediction
• Evaluation results comparing LoRA with other methods like dense layers and the CRISPRon model
• Conclusion highlighting the benefits of using PEFT for biology applications on AWS