Central Limit Theorem for Mutation Systems

TL;DR

This paper establishes a Central Limit Theorem for mutation systems modeling in-vivo DNA data storage, characterizing their stochastic fluctuations and providing a foundation for understanding sequence evolution over time.

Contribution

It introduces a CLT for mutation systems, using spectral analysis and martingale techniques to analyze sequence evolution in DNA storage.

Findings

Derived the asymptotic distribution of sequence counts

Explicitly calculated the limiting covariance matrix

Provided a theoretical framework for error analysis in DNA storage

Abstract

DNA-based storage has emerged as a promising alternative to traditional data storage methods, offering unmatched advantages in data density, longevity, and sustainability. Two main approaches have developed: in-vitro storage, where information is synthesized in controlled environments, and in-vivo storage, where data is embedded within an organism's DNA for enhanced confidentiality and protection. While in-vivo DNA storage provides unique advantages, it faces significant challenges from mutations, including duplications, deletions, and substitutions, which cause sequence evolution over time. Thus, in-vivo systems experience continuous sequence alterations that increase length and change composition, making error correction particularly challenging. We study the asymptotic behavior of mutation systems, which model the probabilistic evolution of sequences over a finite alphabet, and are central to the analysis of in-vivo DNA-based data storage. Building upon prior works that established the limit of empirical $k$-tuple frequencies, we characterize the stochastic fluctuations around these values by establishing a Central Limit Theorem (CLT). Our approach leverages the spectral properties of the $k$-substitution matrix to project the centered count vectors, allowing us to approximate the system via a martingale difference sequence, and then verifying the classical martingale CLT conditions. In addition, we explicitly derive the limiting covariance matrix.