[论文解读] Understanding the Dynamics and Optimizing the Performance of Chemostat Selection Experiments

Aryeh Wides, Ron Milo|arXiv (Cornell University)|Jun 1, 2018

Evolution and Genetic Dynamics参考文献 36被引用 14

一句话总结

本文提出一个定量框架，用于理解并优化微生物进化中的连续培养选择实验，通过解析模型与仿真模拟预测五个实验阶段的动力学行为。研究揭示了反直觉现象（如基质浓度与流入浓度无关），并提出实用的优化准则，以最小化选择时间并确保有效菌株富集，结果表明在最优条件下选择时长具有强鲁棒性。

ABSTRACT

A chemostat enables long-term, continuous, exponential-phase growth in an environment limited as prescribed by the researcher. It is thus a potent tool for laboratory evolution - selecting for strains with desired phenotypes. However, despite the apparently simple design governed by a limited set of rules, analysis of chemostat dynamics shows that they display counter-intuitive properties. For example, the concentration of limiting substrate in the chemostat is independent of the concentration in the influx and only dependent on the dilution rate and the strain parameters. Moreover, choosing optimal operational parameters (dilution rate, volume size, etc.) can be challenging. There are conflicting requirements in the experimental design, such as a need for relatively fast growth conditions for mutation accumulation on the one hand versus slow dilution for a large fitness advantage for mutants to take over the population quickly on the other.In this study, we provide analytic and computational tools to help understand and predict chemostat dynamics, and choose suitable operational parameters. We refer to five stages of the process: (A) parameter choice and setup, (B) basic steady state growth, (C) mutation, (D) single takeover and (E) successive takeovers. We present a qualitative and quantitative framework to answer the questions confronted in each of these stages. We provide a set of simulations which support the quantitative results, and a graphical user interface to give a hands-on opportunity to experience and visualize the analytic results. We detail conditions that produce ineffectual selection regimes, and find that when avoided, the selection time is relatively robust, and usually varies by less than an order of magnitude. Finally, we suggest rules of thumb to help ensure that the chosen parameters lead to effective selection and minimize the duration of the selection process.

研究动机与目标

• 解决连续培养选择实验中阻碍有效实验设计的反直觉动力学行为。
• 识别平衡突变积累与突变体快速取代速度的最优操作参数（稀释率、体积等）。
• 为连续培养选择的五个阶段（准备、稳态、突变、单次取代、连续取代）提供系统性框架。
• 通过避免低效参数区域并确保在各种条件下性能稳健，最小化选择时长。
• 为实验人员提供实用、数据驱动的优化准则，以实现可靠且高效的菌株进化。

提出的方法

• 基于米氏动力学建立解析模型，描述连续培养中基质消耗与种群动力学。
• 通过计算仿真验证并可视化所有五个实验阶段的预测动力学行为。
• 引入图形用户界面（GUI），实现对参数变化对选择结果影响的交互式探索。
• 应用稳态分析，推导稀释率、菌株参数与限制性基质浓度之间的关系。
• 量化稀释率与体积对突变体取代时间与选择效率的影响。
• 识别导致选择无效的条件（如稀释率过高或过低），并推导出稳健的参数范围。

实验结果

研究问题

• RQ1 连续培养中限制性基质浓度如何依赖于实验参数？为何其与流入浓度无关？
• RQ2 何种稀释率与体积设置可使有益突变体在连续培养中最快取代种群？
• RQ3 快速生长以积累突变与缓慢稀释以赋予突变体优势之间的矛盾要求，如何影响选择效率？
• RQ4 在何种条件下选择过程会失效？如何避免这些情况？
• RQ5 当选择最优条件时，选择时长对操作参数变化的鲁棒性如何？

主要发现

• 连续培养中限制性基质浓度仅取决于稀释率与菌株特异性参数，与流入浓度无关。
• 在选择最优条件时，选择时长对参数变化具有相对鲁棒性，通常变化小于一个数量级。
• 选择无效的参数区域（突变体无法取代）通常源于不适当的稀释率，尤其是过高或过低的值。
• 本研究识别出能最大化选择效率并最小化时长的稀释率与体积范围，提供实用的优化准则。
• 仿真结果证实，解析框架能准确预测取代动力学，支持该模型在实验规划中的应用。
• 图形用户界面支持实时探索参数影响，提升实验设计的直觉认知与准确性。