Analysis of chimeric reads characterises the diverse targetome of AGO2-mediated regulation

Abstract

Argonaute proteins are instrumental in regulating RNA stability and translation. AGO2, the major mammalian Argonaute protein, is known to primarily associate with microRNAs, a family of small RNA 'guide' sequences, and identifies its targets primarily via a 'seed' mediated partial complementarity process. Despite numerous studies, a definitive experimental dataset of AGO2 'guide'-'target' interactions remains elusive. Our study employs two experimental methods-AGO2 CLASH and AGO2 eCLIP, to generate thousands of AGO2 target sites verified by chimeric reads. These chimeric reads contain both the AGO2 loaded small RNA 'guide' and the target sequence, providing a robust resource for modeling AGO2 binding preferences. Our novel analysis pipeline reveals thousands of AGO2 target sites driven by microRNAs and a significant number of AGO2 'guides' derived from fragments of other small RNAs such as tRNAs, YRNAs, snoRNAs, rRNAs, and more. We utilize convolutional neural networks to train machine learning models that accurately predict the binding potential for each 'guide' class and experimentally validate several interactions. In conclusion, our comprehensive analysis of the AGO2 targetome broadens our understanding of its 'guide' repertoire and potential function in development and disease. Moreover, we offer practical bioinformatic tools for future experiments and the prediction of AGO2 targets. All data and code from this study are freely available at https://github.com/ML-Bioinfo-CEITEC/HybriDetector/ .

Figure 1

(A) Schematic of Ago2 loaded with a short RNA guide sequence, binding to a target site using a seed driven approach. (B) Schematic of a chimeric read containing fragments of the small RNA guide sequence and the target site. (C) Experimental outline of the CLASH technique. (D) Outline of the bioinformatic pipeline for identification of single genomic, single small RNA, and chimeric reads. (All schematics in this figure were produced using Adobe Illustrator 2023 v27.9).

Figure 2

(A) Read length distribution for sequencing reads annotated as genomic, guide, and chimeric. (B) Fraction of high confidence chimeric interactions per thousand reads for AGO2-CLASH and AGO2-eCLIP experiments. (C) Distribution of identified guide sequences on guide databases. (D) Distribution of genomic target sequences on genic element annotations. (E) Distribution of AGO2-CLASH chimeric reads on guide databases and genic annotation. (F) Distribution of AGO2-eCLIP chimeric reads on guide databases and genic annotations.

Figure 3

(A) Process of chimeric read representation as 2D alignment matrix. (B) Architecture of the Convolutional Neural Network used for training binding prediction models, consisting of three 2D Convolutional blocks followed by a fully connected network. All models were trained on 1:10 imbalanced datasets derived from high quality chimeric interactions. (C) Precision–Recall curve of miRNA trained CNN model against the state of the art, evaluated on left-out balanced high quality chimeric interactions. (D) Precision–Recall curve of tRNA trained CNN model against the state of the art, evaluated on left-out balanced high quality chimeric interactions. (E) Luciferase assay validation of selected chimeric interactions. Interactions with a predicted canonical seed (circle), predicted 3ʹ interaction but no seed (dot), and no clear interaction were select-ed. R/F ratios below 1.0 denote efficient downregulation upon transfection, orange and yellow bars showing significant downregulation within replicates using Student t-test.