AGO2–miRNA complexes were generated by adding synthetic miRNA duplexes to lysate from cells that over-expressed recombinant AGO2, and then these complexes were purified on the basis of affinity to the miRNA seed. RNA libraries were generated by in vitro transcription of synthetic DNA templates. For AGO-RBNS, purified AGO2–miRNA complex was incubated with a large excess of library molecules, and after reaching binding equilibrium, library molecules bound to AGO2–miRNA complex were isolated and prepared for high-throughput sequencing. Examination of k-mers enriched within the bound library sequences identified miRNA target sites, and relative KD values for each of these sites were simultaneously determined by maximum likelihood estimation, fitting to AGO-RBNS results obtained over a 100-fold range in AGO2–miRNA concentration.
Intracellular miRNA-mediated repression was measured by performing RNA-seq on HeLa cells that had been transfected with a synthetic miRNA duplex. For sites that were sufficiently abundant in endogenous 3′ UTRs, efficacy was measured on the basis of their influence on levels of endogenous mRNAs of HeLa cells. Site efficacy was also evaluated using massively parallel reporter assays, which provided information for the rare sites as well as the more abundant ones. The biochemical and biochemical+ models of miRNA-mediated repression were constructed and fit using the measured KD values and the repression of endogenous mRNAs was observed after transfecting miRNAs into HeLa cells. The CNN was built using TensorFlow, trained using the measured KD values and the repression observed in the HeLa transfection experiments, and tested on the repression of endogenous mRNAs observed after transfecting miRNAs into HEK293T cells. Results were also tested on external datasets examining either intracellular binding of miRNAs by CLIP-seq or repression of endogenous mRNAs after miRNAs had been transfected, knocked down, or knocked out. The details of each of these methods are described in thesupplementary materials .
Intracellular miRNA-mediated repression was measured by performing RNA-seq on HeLa cells that had been transfected with a synthetic miRNA duplex. For sites that were sufficiently abundant in endogenous 3′ UTRs, efficacy was measured on the basis of their influence on levels of endogenous mRNAs of HeLa cells. Site efficacy was also evaluated using massively parallel reporter assays, which provided information for the rare sites as well as the more abundant ones. The biochemical and biochemical+ models of miRNA-mediated repression were constructed and fit using the measured KD values and the repression of endogenous mRNAs was observed after transfecting miRNAs into HeLa cells. The CNN was built using TensorFlow, trained using the measured KD values and the repression observed in the HeLa transfection experiments, and tested on the repression of endogenous mRNAs observed after transfecting miRNAs into HEK293T cells. Results were also tested on external datasets examining either intracellular binding of miRNAs by CLIP-seq or repression of endogenous mRNAs after miRNAs had been transfected, knocked down, or knocked out. The details of each of these methods are described in the
