Parikh Saurin Bipin, Castilho Coelho Nelson, Carvunis Anne-Ruxandra
Department of Computational and Systems Biology, Pittsburgh Center for Evolutionary Biology and Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA 15213, USA.
G3 (Bethesda). 2021 Feb 9;11(2). doi: 10.1093/g3journal/jkaa068.
Microbial growth characteristics have long been used to investigate fundamental questions of biology. Colony-based high-throughput screens enable parallel fitness estimation of thousands of individual strains using colony growth as a proxy for fitness. However, fitness estimation is complicated by spatial biases affecting colony growth, including uneven nutrient distribution, agar surface irregularities, and batch effects. Analytical methods that have been developed to correct for these spatial biases rely on the following assumptions: (1) that fitness effects are normally distributed, and (2) that most genetic perturbations lead to minor changes in fitness. Although reasonable for many applications, these assumptions are not always warranted and can limit the ability to detect small fitness effects. Beneficial fitness effects, in particular, are notoriously difficult to detect under these assumptions. Here, we developed the linear interpolation-based detector (LI Detector) framework to enable sensitive colony-based screening without making prior assumptions about the underlying distribution of fitness effects. The LI Detector uses a grid of reference colonies to assign a relative fitness value to every colony on the plate. We show that the LI Detector is effective in correcting for spatial biases and equally sensitive toward increase and decrease in fitness. LI Detector offers a tunable system that allows the user to identify small fitness effects with unprecedented sensitivity and specificity. LI Detector can be utilized to develop and refine gene-gene and gene-environment interaction networks of colony-forming organisms, including yeast, by increasing the range of fitness effects that can be reliably detected.
微生物生长特性长期以来一直被用于研究生物学的基本问题。基于菌落的高通量筛选能够利用菌落生长作为适应性的代理指标，对数千个单菌株进行平行适应性估计。然而，适应性估计因影响菌落生长的空间偏差而变得复杂，这些偏差包括营养物质分布不均、琼脂表面不规则以及批次效应。为校正这些空间偏差而开发的分析方法依赖于以下假设：（1）适应性效应呈正态分布，以及（2）大多数基因扰动导致适应性的微小变化。尽管这些假设在许多应用中是合理的，但并不总是成立，并且可能限制检测微小适应性效应的能力。特别是在这些假设下，有益的适应性效应 notoriously difficult to detect（此处疑为拼写错误，可能是“ notoriously difficult to detect”，意为“众所周知难以检测”）。在这里，我们开发了基于线性插值的检测器（LI Detector）框架，以实现基于菌落的灵敏筛选，而无需对适应性效应的潜在分布做出先验假设。LI Detector使用参考菌落网格为平板上的每个菌落分配一个相对适应性值。我们表明，LI Detector在校正空间偏差方面是有效的，并且对适应性的增加和减少同样敏感。LI Detector提供了一个可调系统，允许用户以前所未有的灵敏度和特异性识别微小的适应性效应。通过扩大能够可靠检测的适应性效应范围，LI Detector可用于开发和完善包括酵母在内的菌落形成生物体的基因-基因和基因-环境相互作用网络。
