Paper List

期刊: ArXiv Preprint
发布日期: 2025-12-01
BioinformaticsComputational Biology

MoRSAIK: Sequence Motif Reactor Simulation, Analysis and Inference Kit in Python

Max-Planck-Institut für Astrophysik | Ludwig-Maximilians-Universität München | Technische Universität München | Exzellenzcluster ORIGINS

Johannes Harth-Kitzerow, Ulrich Gerland, Torsten A. Enßlin

30秒速读

IN SHORT: This work addresses the computational bottleneck in simulating prebiotic RNA reactor dynamics by developing a Python package that tracks sequence motif concentrations instead of full RNA strands, enabling efficient Bayesian inference of reaction parameters.

核心创新

  • Methodology First implementation of Bayesian inference methods for RNA reactor simulations using Geometric Variational Inference via NIFTy.re framework
  • Methodology Novel mean-field approximation approach that tracks k-mer motif concentrations (default k=4) instead of exponentially growing full RNA sequences
  • Biology Enables systematic investigation of templated ligation dynamics under varying environmental conditions relevant to RNA world hypothesis

主要结论

  • MoRSAIK reduces computational complexity from exponential to polynomial by tracking k-mer motifs (k=4 default) instead of full RNA strands
  • The package enables Bayesian inference of reaction rate constants from templated ligation count data using Geometric Variational Inference
  • Integration with JAX provides differentiable models for efficient gradient-based optimization and uncertainty quantification
研究空白： Current RNA reactor simulations face exponential computational complexity with strand length, making parameter inference from experimental data impractical for realistic sequence spaces.

摘要: Origins of life research investigates how life could emerge from prebiotic chemistry only. Living systems as we know them today rely on RNA, DNA and proteins. According to the central dogma of molecular biology, information is stored in DNA, transfered by RNA resulting in proteins that catalyze functional reactions, such as synthesis and replication of DNA and RNA. One possible explanation of how this mechanism evolved provides the RNA world hypothesis (Crick 1968; Higgs and Lehman 2014; Orgel 1968; Pressman, Blanco, and Chen 2015; Szostak 2012). It states that life could emerge from RNA strands only, storing and transferring biological information, as well as catalyzing reactions as ribozymes. Before this state could have emerged, however, the prebiotic world was probably a purely chemical pool of short RNA strands with random sequences and without biological function. Despite the lack of guidence by proteins, the RNA sequences reacted with each other. In such an RNA reactor RNA strands perform hybridization and dehybridization, as well as ligation and cleavage. In this context relevant questions are what are the conditions that allow longer RNA strands to be built and how can information carrying in RNA sequence emerge? A key reaction for the emergence of longer RNA strands is templated ligation. There, two strands hybridize adjacent onto a template strand and ligate. The rate of this reaction is the larger, the better the two strands match the complementary sequence of the template strand. The extended strands can then serve as a template for the next generation of templated ligation. This leads to an acceleration of production of complementary strands. This process, however, is highly sensitive to environmental conditions determining the reaction rates within an RNA reactor (Göppel et al. 2022; Rosenberger et al. 2021). In order to investigate those RNA reactors, efficient simulations are needed because the space of possible RNA sequences increases exponentially with the length of the strands, as well as the number of reactions between two strands. In addition, simulations have to be compared to experimental data for validation and parameter calibration. Here, we present the MoRSAIK python package for sequence motif (or k-mer) reactor simulation, analysis and inference. It enables users to simulate RNA sequence motif dynamics in the mean field approximation as well as to infer the reaction parameters from data with Bayesian methods and to analyze results by computing observables and plotting. MoRSAIK simulates an RNA reactor by following the reactions and the concentrations of all strands inside up to a certain length (of four nucleotides by default). Longer strands are followed indirectly, by tracking the concentrations of their containing sequence motifs of that maximum length.