Technology explainer
How AI-designed RNA is validated in the laboratory
From computational structure proposals to synthesis, biochemical assays and iterative redesign.
An AI model can propose an RNA backbone and nucleotide sequence, but a computational score does not prove that the molecule folds or functions in the laboratory.
Design and synthesis
Researchers first specify a function and structural motifs that must be preserved. Models generate candidates, which are filtered for geometric plausibility and sequence compatibility. Selected sequences are then chemically synthesized or transcribed.
Functional assays
The experiment must match the claim. A ribozyme can be tested by measuring how much substrate it cleaves under controlled conditions. A fluorogenic aptamer can be evaluated by its fluorescence after binding a target. Controls include natural molecules, expert designs and inactive variants.
Iteration and independence
Experimental results can guide another design round, but reporting only the best candidates can exaggerate reliability. A strong evaluation reports the total success rate, uses identical resource limits for competing methods and asks independent laboratories to reproduce the findings.
First appeared in
RhoAgent Designed Active RNA, Including a Sixfold Ribozyme Result