基于特征挖掘和机器学习的细菌VI型分泌系统效应分子的功能分类、计算预测和实验验证

Many pathogens use their type VI secretion systems (T6SS) to inject various effector proteins into target cells, which is important for their virulence or environmental survivability. Recent researches reveal that T6SS effectors play three functional roles by targeting to three different subcellular locations. And the congenetic immunity proteins of these effectors can protect their hosts from targeted attacks. How to identify novel T6SS effectors in genomic proteins is a challenge with the accumulation of the sequenced bacterial genomic data. The absence of accurate bioinformatics prediction methods is keeping this problem unsolved. In our previous studies, we found a conserved Ultra-MIX motif widely existed in the N-terminal sequences of T6SS effectors. By establishing a hidden Markov model for this motif, we performed a large-scale search in experimental-validated effectors secreted by T3SS, T4SS and T6SS, and also in proteins from dozens of pathogenic genomes. The results suggested that Ultra-MIX is possibly a particular motif in T6SS effectors. We also found many feature divergences between the different types of effectors. Based on these results, we deduce that computational identification of effectors and their functional types is feasible if various discriminating features of effectors are clear enough. As such this study focuses on the functional classification, computational prediction and experimental validation of T6SS effectors. Firstly, we will compute the effector characteristics using feature mining techniques, build prediction models using machine learning techniques, and develop prediction algorithms and online tools. Then, we will perform experimental identification of predicted effector candidates in Pseudomonas aeruginosa. Through this study, several effective tools will be developed for genomic T6SS effector prediction, effector classification of functional types, and genomic immunity protein detection. Furthermore, some novel effectors secreted by T6SS will be discovered, which may reveal new ways of Pseudomonas aeruginosa interacting with target cells via its T6SS.

病原菌利用VI型分泌系统(T6SS)分泌效应分子，增强其致病性或生存能力。最新研究表明这些效应分子分属不同的功能类型。随着细菌基因组数据大量增长，如何快速鉴定其中的效应分子成为亟待解决的问题。在前期研究中，课题组发现了效应分子序列中的一种保守模式，计算了该模式的分布特征，推断该模式为T6SS效应分子所特有，此外还发现了不同功能效应分子间的多种特征差异。若能进一步明确这些特征，则可以通过计算的方法预测这类分子及其功能，加速实验鉴定进程。为此，本研究将利用课题组掌握的生物信息学技术，挖掘这类分子的未知特征，设计合适的机器学习模型，建立能够预测T6SS效应分子、识别效应分子功能和检测其免疫蛋白的多种算法，并在铜绿假单胞菌中进行实验验证。通过本研究，可望为相关人员提供用于T6SS效应分子预测和功能分类的有效工具，证实铜绿假单胞菌基因组内新的效应分子，揭示该类细菌通过T6SS攻击靶细胞的新途径。

细菌通过VI型分泌系统（T6SS）分泌效应分子，实现对宿主细胞的侵入，或者攻击环境中的竞争对手，维持自身的生存和繁殖。细菌基因组内存在哪些T6SS效应分子，发挥怎样的功能，都需要有快速有效的工具来进行鉴定。本项目筛选T6SS效应分子数据集，分析效应分子的序列和进化特征，设计机器学习预测模型，开发相应的算法和工具，在细菌基因组内实现对T6SS效应分子的准确预测。并且，对实验室收集的铜绿假单胞菌菌株进行全基因组测序，对基因组数据进行预测和分析，揭示这种细菌的T6SS特征，筛选候选的效应分子来进行实验鉴定。此外，我们进一步探索T6SS在肠道菌群中的特征和潜在的功能，建立了肠道菌群的T6SS组分的分析方法。本研究建立预测方法，可以为同行提供准确的生物信息学筛选工具，满足实验前快速筛选候选效应分子的需求，并且，能够阐明和揭示T6SS效应分子的一些关键特征，为效应分子的功能研究提供有价值的结论。本研究的成果发表在Briefings in Bioinformatics、Frontiers in Microbiology等专业期刊上，并且建立了免费开放的标准测试数据集和Web预测服务器。