Additional file 1 of Longitudinal host-microbiome dynamics of metatranscription identify hallmarks of progression in periodontitis

Additional file 1. Figure S1.Experimental design. 15 participants were selected from a total cohort of 415 participants. These patients had the clinal conditions we wanted: progressing sites where CAL increased steadily and significantly during the study and stable sites where CAL values remained significantly identical. Genetic background should have a minimal effect on the outcome of individual sites. At baseline, all teeth used were clinically identical. Subgingival plaque samples were taken every 2 months for 1 year, after which all patients underwent scaling and root planing as treatment. After 3 months for a visual check-up and again after 6 months when they were monitored, all returned to the clinic, and samples were also taken. Figure S2. Phylogenetic assignment and relative quantification of microbiome metatranscriptome. a) Combined clusters with peaks before the change point. b) Venn diagram of the combined species from the stable and progressing clusters. c) List of the species in the three sections of the Venn diagram. Figure S3. Gene set enrichment analysis of the overall changes in the study. We performed gene set enrichment analysis of GO terms and KEGG pathways using clusterProfiler [49] on the DE from the host and the microbiome. In red activated enriched activities. In green suppressed enriched activities. Figure S4. Microbiome GO ontology and KEGG pathways enrichment analysis of differentially expressed (DE) gene clusters in stable sites. Clusters of DE genes were obtained by determining an optimal number of clusters using fviz_nbclust from the ‘factoextra’ package with the gap statistic method and performing clustering using tsclust with shape-based distance (SBD), which makes the clustering particularly useful for time series where the shape matters more than exact numerical values. Clusters are standardized to log2 fold-change of abundance. Colors and cluster numbers are arbitrary. The actual composition of the different clusters is presented in Table S2. Enrichment of gene sets was performed using the Cytoscape app ClueGO with the GO biological process, KEGG pathways, and KEGG-human disease ontologies. In red, metabolic activities were activated (enriched), and in blue, metabolic activities were repressed. Figure S5. Microbiome GO ontology and KEGG pathways enrichment analysis of differentially expressed (DE) gene clusters in progressing sites. Clusters of DE genes were obtained by determining an optimal number of clusters using fviz_nbclust from the ‘factoextra’ package with the gap statistic method and performing clustering using tsclust with shape-based distance (SBD), which makes the clustering particularly useful for time series where the shape matters more than exact numerical values. Clusters are standardized to log2 fold-change of abundance. Colors and cluster numbers are arbitrary. The actual composition of the different clusters is presented in Table S5. Enrichment of gene sets was performed using the Cytoscape app ClueGO with the GO biological process, KEGG pathways, and KEGG-human disease ontologies. In red, metabolic activities were activated (enriched), and in blue, metabolic activities were repressed. Figure S6. CAL-host and microbiome delay correlation analysis, GO ontology, and KEGG pathways enrichment analysis in stable sites. Using the R package dynOmics [73] we measured the correlation between the two time-series (CAL and host or microbiome genes) at different time lags. We identified the delay (positive or negative lag) at which two variables correlate most strongly. a) CAL profile preceded human activities by 2 months. b) Microbiome activities that preceded CAL by 2 months. c) CAL profile preceded microbiome activities by 2 months. The percentages represent the %terms/group, that is, the proportion of terms within each functional group or category relative to the total number of terms in your analysis. It shows how many enriched terms are associated with each functional group, showing their relative importance. n = number of nodes in the network. In red, metabolic activities were activated (enriched), and in blue, suppressed activities were suppressed. Figure S7. Host and microbiome-CAL delay correlation analysis, GO ontology, and KEGG pathways enrichment analysis in progressing sites. Using the R package dynOmics [73] we measured the correlation between the two time-series (CAL and host or microbiome genes) at different time lags. We identified the delay (positive or negative lag) at which two variables correlate most strongly. a) Human activities that preceded CAL by 2 months. b) Microbiome activities that preceded CAL by 2 months. The percentages represent the %terms/group, that is, the proportion of terms within each functional group or category relative to the total number of terms in your analysis. It shows how many enriched terms are associated with each functional group, showing their relative importance. n = number of nodes in the network. In red, metabolic activities were activated (enriched), and in blue, suppressed activities were suppressed. Figure S8. CAL-host and microbiome delay correlation analysis, GO ontology, and KEGG pathways enrichment analysis in progressing sites. Using the R package dynOmics [73] we measured the correlation between the two time-series (CAL and host or microbiome genes) at different time lags. We identified the delay (positive or negative lag) at which two variables correlate most strongly. a) CAL profile preceded human activities by 2 months. b) CAL profile preceded microbiome activities by 2 months. The percentages represent the %terms/group, that is, the proportion of terms within each functional group or category relative to the total number of terms in your analysis. It shows how many enriched terms are associated with each functional group, showing their relative importance. n = number of nodes in the network. In red, metabolic activities were activated (enriched), and in blue, suppressed activities were suppressed.

补充文件1。图S1：实验设计。从共计415名受试者的队列中筛选出15名参与者。这些患者符合我们研究所需的临床特征：存在研究期间临床附着水平（Clinical Attachment Level，CAL）持续且显著升高的进展位点，以及CAL值始终无显著变化的稳定位点。个体位点的结局受遗传背景的影响应极小。基线时，所有受试牙齿的临床状态均一致。在为期1年的随访周期内，每2个月采集一次龈下菌斑样本；随访结束后，所有患者均接受龈下刮治与根面平整治疗。分别于治疗后3个月进行视觉复查，以及治疗后6个月开展监测时，所有患者均返回诊所并再次采集样本。图S2：微生物组宏转录组的系统发育分类与相对定量。a）变化点前的峰合并聚类结果；b）稳定位点与进展位点聚类合并得到的物种维恩图；c）维恩图三个分区中的物种列表。图S3：研究整体变化的基因集富集分析。我们针对宿主与微生物组的差异表达（Differentially Expressed，DE）基因，使用clusterProfiler工具[49]对基因本体（Gene Ontology，GO）术语与京都基因与基因组百科全书（Kyoto Encyclopedia of Genes and Genomes，KEGG）通路进行基因集富集分析。红色代表激活的富集活性，绿色代表抑制的富集活性。图S4：稳定位点差异表达（DE）基因簇的微生物组GO本体与KEGG通路富集分析。差异表达基因簇的获取方式为：使用‘factoextra’包中的fviz_nbclust函数结合间隙统计量法确定最优聚类数，再采用基于形状的距离（Shape-based Distance，SBD）的tsclust函数进行聚类，该聚类方法尤其适用于形状特征较精确数值更重要的时间序列数据。所有聚类结果均标准化为丰度的log₂倍变化值。聚类颜色与编号均为任意设定，不同聚类的实际组成详见表S2。基因集富集分析通过Cytoscape插件ClueGO完成，涵盖GO生物过程、KEGG通路及KEGG人类疾病本体。红色代表代谢活性被激活（富集），蓝色代表代谢活性被抑制。图S5：进展位点差异表达（DE）基因簇的微生物组GO本体与KEGG通路富集分析。差异表达基因簇的获取方式为：使用‘factoextra’包中的fviz_nbclust函数结合间隙统计量法确定最优聚类数，再采用基于形状的距离（SBD）的tsclust函数进行聚类，该聚类方法尤其适用于形状特征较精确数值更重要的时间序列数据。所有聚类结果均标准化为丰度的log₂倍变化值。聚类颜色与编号均为任意设定，不同聚类的实际组成详见表S5。基因集富集分析通过Cytoscape插件ClueGO完成，涵盖GO生物过程、KEGG通路及KEGG人类疾病本体。红色代表代谢活性被激活（富集），蓝色代表代谢活性被抑制。图S6：稳定位点的CAL-宿主与微生物组延迟相关性分析，以及GO本体与KEGG通路富集分析。我们使用R包dynOmics[73]，针对不同时间延迟下的两组时间序列数据（CAL与宿主基因或微生物组基因）计算相关性，并识别出两个变量相关性最强的延迟时间（正延迟或负延迟）。a）CAL变化超前于宿主活性2个月；b）微生物组活性超前于CAL2个月；c）CAL变化超前于微生物组活性2个月。图中百分比代表「组内术语占比」，即每个功能组或类别中的术语数量占分析总术语数的比例，用于展示每个功能组关联的富集术语数量，体现其相对重要性。n代表网络中的节点数。红色代表代谢活性被激活（富集），蓝色代表代谢活性被抑制。图S7：进展位点的宿主与微生物组-CAL延迟相关性分析，以及GO本体与KEGG通路富集分析。我们使用R包dynOmics[73]，针对不同时间延迟下的两组时间序列数据（CAL与宿主基因或微生物组基因）计算相关性，并识别出两个变量相关性最强的延迟时间（正延迟或负延迟）。a）宿主活性超前于CAL2个月；b）微生物组活性超前于CAL2个月。图中百分比代表「组内术语占比」，即每个功能组或类别中的术语数量占分析总术语数的比例，用于展示每个功能组关联的富集术语数量，体现其相对重要性。n代表网络中的节点数。红色代表代谢活性被激活（富集），蓝色代表代谢活性被抑制。图S8：进展位点的CAL-宿主与微生物组延迟相关性分析，以及GO本体与KEGG通路富集分析。我们使用R包dynOmics[73]，针对不同时间延迟下的两组时间序列数据（CAL与宿主基因或微生物组基因）计算相关性，并识别出两个变量相关性最强的延迟时间（正延迟或负延迟）。a）CAL变化超前于宿主活性2个月；b）CAL变化超前于微生物组活性2个月。图中百分比代表「组内术语占比」，即每个功能组或类别中的术语数量占分析总术语数的比例，用于展示每个功能组关联的富集术语数量，体现其相对重要性。n代表网络中的节点数。红色代表代谢活性被激活（富集），蓝色代表代谢活性被抑制。