ASOkai - An open-source framework for ASO design using kinetics and AI

Poster presented on Oct. 20, 2025 by Nathalie Gocht at 21st Annual Meeting of the Oligonucleotide Therapeutics Society, Budapest, Hungary.

Abstract: The importance of in silico drug design in light of the enormous combinatorial design space for Antisense Oligonucleotides (ASOs) is reflected in the advent of tools such as PFRED (siRNAs and ASO design), eSkip-Finder (a machine learning (ML)-based web app for exon skipping ASOs), and ASOptimizer (ML-system for ASOs targeting IDO1 with optimized modifications). Common to these frameworks is the use of sequence features as well as chemical and thermodynamic features to predict and evaluate the potential efficacy of drug candidates. Our open-source ASO design pipeline emphasizes aspects of ASO design which have not been routinely utilized or required human experts. In particular our focus is on comprehensive analysis also of unspecific off-targets and on ASO kinetics, for targets, specific and unspecific off-targets, currently for RNase H1 mediated knock-down. The kinetic analyses are based on extensions of the kinetic model introduced in Pedersen et al. ; AI models allow prediction of stochastic simulation outcomes and thus consideration of kinetics already in the initial screening phase; interfaces to standard stochastic simulation packages facilitate running detailed simulations. AI-models are also used to avoid the exponential computing cost for identifying low-affinity specific off-target and statistics of even lower-affinity, unspecific off-target binding sites. The design of the software allows use on laptops and within workflow system; individual computational steps can be used independently and be included in in-house systems. By providing an open-source basis for further developments by academic and industry groups we hope that it will serve as a crystallization point for community-building through open exchange of knowledge.

Download PDF of Poster.