Bioinformatics: Novel Machine Learning Approach Developed by Researchers

In the ongoing battle against viruses, bacteria, and other pathogens, synthetic biology presents innovative technological avenues, the effectiveness of which is being substantiated through experimental validation. Scientists from the Würzburg Helmholtz Institute for RNA-based Infection Research, in collaboration with the Helmholtz AI Cooperative, have harnessed the power of data integration and artificial intelligence (AI) to devise a machine learning methodology. This approach promises more precise predictions regarding the efficacy of CRISPR technologies compared to previous methods.

Their findings have been detailed in the esteemed journal Genome Biology.

At the core of every organism lies its genome or DNA, serving as the blueprint for proteins and governing cellular replication. With the aim of combating pathogens, addressing genetic ailments, and other beneficial pursuits, molecular biological CRISPR technologies have emerged as tools to selectively modify or suppress genes and impede protein synthesis.

Among these tools is CRISPRi (CRISPR interference), which interferes with gene expression without altering the DNA sequence. Unlike the CRISPR-Cas system, commonly referred to as “gene scissors,” CRISPRi employs a ribonucleic acid (RNA) as a guide to direct a nuclease (Cas), which binds to the DNA without cleaving it. This binding prevents the transcription of the target gene, rendering it inactive.

However, accurately predicting the performance of CRISPRi for specific genes has posed a challenge until now. Collaborating with the University of Würzburg and the Helmholtz Artificial Intelligence Cooperation Unit (Helmholtz AI), researchers from the Würzburg Helmholtz Institute for RNA-based Infection Research have introduced a novel machine learning approach, leveraging data integration and artificial intelligence (AI) to enhance predictive capabilities.

The Methodology: CRISPRi screens are highly sensitive tools used to evaluate the effects of reduced gene expression. In their research, scientists utilized data from multiple genome-wide CRISPRi essentiality screens to train their machine learning model. Their objective was to refine predictions regarding the effectiveness of the engineered guide RNAs employed in the CRISPRi system.

Lars Barquist, a computational biologist leading a bioinformatics research group at the Würzburg Helmholtz Institute, elaborates, “Genome-wide screens offer only indirect insights into guide efficiency. Therefore, we have employed a novel machine learning technique to discern the efficacy of guide RNAs from the influence of silenced genes.”

Assisted by additional AI tools (“Explainable AI”), the team established transparent design principles for future CRISPRi experiments. Validation of their approach involved conducting an independent screen targeting essential bacterial genes, demonstrating superior predictive accuracy compared to existing methodologies.

Yanying Yu, a Ph.D. student in Lars Barquist’s research group and the study’s lead author, affirms, “Our model has exhibited superior performance over existing methods, offering more dependable forecasts of CRISPRi efficacy when targeting specific genes.”

Unexpectedly, the researchers discovered that the guide RNA itself is not the primary determinant of CRISPRi effectiveness in essentiality screens. Yu elucidates, “Certain gene-specific characteristics related to gene expression appear to exert a greater influence than previously thought.”

Furthermore, the study underscores the significance of integrating data from multiple sources in enhancing predictive accuracy, enabling a more robust evaluation of guide RNA efficiency. Barquist emphasizes, “Augmenting our training data through the amalgamation of diverse experiments is crucial for developing more accurate prediction models. Prior to our study, insufficient data posed a significant impediment to prediction accuracy.”

The methodology outlined in their publication promises to facilitate the planning of more effective CRISPRi experiments in the future, benefiting both biotechnology and fundamental research endeavors. Barquist concludes, “Our study offers a framework for the development of precision tools to manipulate bacterial gene expression, ultimately aiding in the comprehension and mitigation of pathogens.”