3.1 Data statistics and analysis types provided by RAD-Blood
The statistics of data types and analysis types in RAD-Blood are illustrated in Fig. 1. Three types of RNA-seq data of AD blood were included in this study, including bulk mRNA-seq, miRNA-seq, and scRNA-seq, with 1302, 147, and 24 samples, respectively. These data were then processed, analyzed, and visualized through a uniform pipeline. The results were presented in the website of RAD-Blood (http://www.bioinform.cn/RAD-Blood/). For all of the data types, differential expression analysis and pathway enrichment analysis were performed. Immune cell abundance analysis and TCR analysis were specifically conducted for bulk mRNA-seq data. In addition, cell marker, cell annotation, and intercellular communication analysis were designed for scRNA-seq data.
3.2 Operating procedure of RAD-Blood
The operation procedure of RAD-Blood is illustrated in Fig. 2. Firstly, users can click on the sidebar panel to access the mRNA-seq, miRNA-seq, and scRNA-seq modules (Fig. 2A). Then, they can select a data set with specific species and sample size, and select other arguments like “condition” and “GSEA type” from the drop down lists (Fig. 2B). For example, by clicking the “scRNA-seq” button on the sidebar panel, users can get five sub-modules (Fig. 2C), including “Cell annotation”, “Cell marker”, “Comparison between cell types”, “DEGs”, and “Cell communication”. By clicking the “Cell communication” panel for example, users can obtain the cell communication results, including “Cell interaction”, “Function and structure”, “Signal flow”, “L-R pairs”, and “Signal pathway”, from the selected scRNA-seq data set. Furthermore, by clicking the “Cell interaction” sub-panel from the “Cell communication” panel (Fig. 2C), a table is showing the detailed information about communications between cell types, a histogram showing the statistics of the number and the strength of interactions between cell types for each sample group, circle plots showing the interaction network among all cell types in each sample group, and scatter diagrams presenting the comparison of outgoing and incoming interaction strength of each cell type in 2D space (Fig. 2D).
3.3 Functions provided by RAD-Blood
RAD-Blood provides three functional modules, including mRNA-seq, miRNA-seq, and scRNA-seq of AD blood. The web structure and typical output mode are shown in Fig. 3. Detailed functions for each module are described as follows.
3.3.1 mRNA-seq
The mRNA-seq module consists of five sub-modules (Fig. 3A): 1) DEG sub-module calculates the DEGs between AD/MCI and normal; 2) GSEA sub-module performs pathway enrichment analysis for the DEGs; 3) Immune abundance sub-module estimates the abundance of immune cells and compares it between AD/MCI and normal; 4) Immune & Expression sub-module calculates the correlation between gene expression and immune cell abundance; 5) TCR sub-module performs TCR analysis and calculates the usage of V(D)J genes and CDR3 amino acid in AD/MCI and normal samples.
3.3.2 miRNA-seq
The miRNA-seq module provides three sub-modules (Fig. 3B): 1) DEG sub-module calculates the differentially expressed miRNAs; 2) “GSEA in miEAA” and 3) “GSEA in mirPath” sub-modules conduct pathway enrichment analysis of DEGs by miEAA[29] and mirPath[30] databases, respectively.
3.3.3 scRNA-seq
The scRNA-seq module provides five sub-modules (Fig. 3C): 1) Cell annotation sub-module annotates cell types (CTs); 2) Cell marker sub-module calculates the CT specific expressed genes and performs GSEA; 3) Comparison between cell types sub-module calculates the DEGs between one CT and another CT, and performs GSEA; 4) DEG sub-module calculates the DEGs for each CT between AD/MCI and normal samples; 5) Cell communication sub-module estimates the strength and number of interactions between cell types.
3.4 Case study
RAD-Blood is a platform for RNA-seq data analysis from AD blood, which is flexible, dynamic, and easy to use. To illustrate the powerful ability of RAD-Blood in finding AD blood bio-markers, we used RAD-Blood to perform a case study. Based on the integrated data set, which integrates SRP310421, ACOM, Emory, and MCSA data sets with batch effects removed (Figure S1), we investigated the expression profiles of mRNAs for AD, MCI, and normal samples. We detected 45174 mRNAs in all samples, of which nearly 17.5% were expressed with TPM > 10, and 52.6% were expressed at a low expression level (TPM: 0–1; Figs. 4A). Among mRNAs transcript from protein coding genes, differential expression analysis was performed and the overlapped DEGs among AD vs. control, AD vs. MCI, and MCI vs. control were shown in Fig. 4B. Among the DEGs, 256 (5.3%) were consistently increased and 18 (0.2%) were consistently decreased from control to MCI to AD, indicating that these mRNAs were potential biomarkers associated with AD progression. The expression fold changes of the top 50 up-regulated genes and the 18 down-regulated genes were shown in Fig. 4C.
Then, we investigated the scRNA-seq data of AD blood using the RAD-Blood platform by selecting the SRP309935 data set (whole blood from 3 AD patients and 2 normal people). Notably, among the 18 consistently down-regulated genes (Fig. 4B), four genes were also the cell markers of the erythroid cell and were down-regulated in erythroid cells from AD blood (Figure S2A). The four genes were HBB, HBD, HBA1, and HBA2, which were predominantly over-expressed in the erythroid cell (Fig. 5A, 5B, and S2B). While these four genes were down-regulated in blood erythroid cells from AD patients compared with that from normal people (Fig. 5C). Interestingly, the proportion of the erythroid cell in the blood of AD patients was also lower than that in normal people (Figure S2C). These results suggest that the decreased proportion and the down-regulated cell markers (hemoglobin subunit beta (HBB), HB delta (HBD), HB alpha 1 (HBA1), and HBA2) of erythroid cells could be indicators of AD, which has not been revealed by existing researches.
To further explore the functional changes of erythroid cells in AD blood, we performed cell communication analysis. Although the total number of interactions in the blood of AD patients was lower than that of normal people (Figure S2D), many weak interactions between erythrocytes and other cell types were observed in the blood of AD patients (Fig. 6A) compared with that of normal people. Also, the interactions between erythroid cells and natural killer cells (cluster 1) or plasmacytoid dendritic cells (cluster 2) in the blood of AD patients were stronger than that of normal people (Fig. 6A). The newly observed signals of erythroid cells in AD blood include outgoing and incoming signals (Fig. 6B). The AD’s unique incoming signals were mainly conducted by CD22, amyloid precursor protein (APP), GALECTIN, and c-type lectin receptors (CLEC) pathways (Fig. 6B). The AD’s unique outgoing signals of the erythroid cells are mainly conducted by MHC-I, ANNEXIN, CD45, LCK, integrin subunit beta 2 (ITGB2), and CLEC pathways (Fig. 6B). The outgoing signals source from erythroid cells to CD8 + T cells and plasmacytoid dendritic cells were transferred through the pairing of human leukocyte antigen (HLA) class I (HLA-A, HLA-B, HLA-C, and HLA-E) ligands with CD8B and CD8A receptors in MHC-I pathway (Fig. 6C). Unsurprisingly, HLA-A, HLA-B, HLA-C, and HLA-E were up-regulated in blood erythroid cells of AD patients compared with normal people (Fig. 6D). Overall, the case study indicates that RAD-Blood is a solution to find potential novel signatures in AD blood.