For constructing high-expression gene sets, measuring the expression level of genes in different cell types is necessary. Transcriptomic techniques, e.g. bulk and single-cell RNA-seq, are the main methods to measure gene expression levels in tissues or cells currently (Stark et al., 2019). In our study, human RNA-seq datasets of different cell types (shown in Table 1 ) from the NCBI GEO database (https://www.ncbi.nlm.nih.gov/geo/ ) were selected for constructing corresponding datasets. The detailed process for selecting datasets is mentioned in Supplemental Methods (Li et al., 2009; Liao et al., 2014; Kim et al., 2019). Finally, 19 bulk RNA-seq datasets (including 2 datasets with corresponding Ribo-seq data) corresponding to different types of cells and 2 single-cell RNA-seq datasets corresponding to major target ‘organs’ of SARS-CoV-2 and HIV-1 (lung and PBMC respectively) were selected.

To determine gene expression levels by RNA-seq, a series of processes are needed to convert raw sequencing data into a normalized gene-level expression matrix (FPKM). The pipeline of this process is mentioned in Supplemental Methods. (Li et al., 2009; Liao et al., 2014; []Kim et al., 2019) For single-cell RNA-seq data, the R package Seurat (version: 4.1.1) was used to preprocess, cluster and visualize cells according to their gene expression patterns. Cell types were annotated according to the corresponding literature of the dataset. Only cell types with more than ten cells in each single-cell RNA-seq dataset were used for CAI calculation and further analysis.

Gene-level expression data (raw counts and FPKM) for all RNA-seq datasets used in this study is provided in Github (https://github.com/Renruohan/CAIvirus). For two single-cell RNA-seq datasets, the corresponding relationship between cells and cell types/large groups was also listed in Github. Also, intermediate results of single-cell RNA-seq datasets by Seurat including cell clustering and visualization in low-dimension space, ratio or number of cells belonging to each cell type/large group and markers to identify each cell type/large group were shown in Supplemental Figures 1 , 2 .

According to the concept of CAI, the definition of the high-expression gene set is arbitrary. In previous studies, typical housekeeping genes like ribosomal genes and histone genes are usually utilized to construct the high-expression gene set. The lack of cell type-specificity made this strategy unsuitable for our study. For highlighting differences in gene expression patterns in various cell types, we selected the top 200 protein-coding genes (contain both housekeeping genes and genes with tissue/cell-specific expression) with the highest mean FPKM level in biological repetitions (for bulk RNA-seq datasets) or cells (for single-cell RNA-seq datasets) for a cell type to construct the high-expression gene set of this cell type. Also, we found that expression levels of some genes changed significantly in different biological repetitions or cells of the same cell type. Hence, before selecting high-expression genes as mentioned above, we removed genes whose standard deviations of the expression levels are large than the average value. The selected top 200 high-expression gene sets and their normalized expression level (FPKM) for all RNA-seq datasets used in this study are provided in Supplemental File 1 . To characterize the functional characteristics of high-expression gene sets in different cell types, we performed GO-BP and KEGG enrichment analysis on them by R package clusterProfiler (Wu et al., 2021) (version: 4.0.5) and Org.Hs.eg.db (version: 3.13.0). Results of enrichment analysis for three representative datasets (bloodmonocytes, CD4+ Tfh in bulk RNA-seq datasets and dendritic cells in scRNA-seq datasets) are provided in Supplemental File 2 .

Different from prokaryotic genes, mRNAs transcripted from eukaryotic genes usually experience complex processing including splicing. Because of the existence of alternative splicing, the same mRNA precursor can be processed to different mature mRNAs and translated to different protein products. As a result, it is necessary to discriminate different splicing isoforms of genes in the high-expression gene set. The process of calculating the approximate expression level of splicing isoforms of high-expression genes is mentioned in Supplemental Methods. For each gene, only the isoform with the highest approximate expression level of each gene was utilized to construct the codon usage frequency table of the high-expression gene set, then p_k and q_k in this expression of CAI could be acquired from this table. CDS sequences of the isoform with the highest approximate FPKM of each gene in high-expression gene sets for all RNA-seq datasets used in this study are provided in Supplemental File 3 .

For verifying if CAI can represent genes’ translational efficiency, we calculated the CAI of the top 5000 highly expressed genes (except genes that are not detected in Ribo-seq experiments) in two groups of paired RNA-seq and Ribo-seq datasets of control cells and cells infected with HIV-1 or SARS-CoV-2 for a certain time. Also, we calculated the translational efficiency by dividing their normalized expression values (FPKM) in Ribo-seq datasets by values in RNA-seq datasets of these genes and then took the logarithm. Then we performed a linear regression analysis between genes’ CAI and translational efficiency in each cell type. Furthermore, for verifying if the CAI of endogenous genes calculated by cell type-specific background gene sets correlated better with translational efficiency than CAI based on commonly used nonspecific background gene sets, we constructed a nonspecific background gene set reference to previously mentioned methods. First, typical housekeeping genes-ribosomal protein genes were selected. Then, we filtered them according to expression patterns, preserving only the ribosomal genes that were present in the top 200 highly expressed genes of at least one cell type (except cells infected by virus or stimulated by cytokines and paired control groups in the same dataset). The CAI of endogenous genes based on this nonspecific background gene set and the correlation between this ‘nonspecific’ CAI and translational efficiency is calculated in the same way as above. In CAI calculation, for each gene in the nonspecific background gene set, one isoform from all isoforms that appear in the top 200 highly expressed gene set of at least one cell type was selected randomly and utilized in the construction of the codon usage frequency table.

Data processing for Ribo-seq datasets is mentioned in Supplemental Methods (Dobin et al., 2012).

Gene-level expression data (raw counts and FPKM) for all Ribo-seq datasets used in this study is provided on the Github page. The CAIs and the translational efficiency of the top 5000 highly-expressed genes for all RNA-seq/Ribo-seq pair datasets in this study are provided in Supplemental File 4 .

In this section, we mainly focus on the CAIs of HIV-1 in different human cell types. We calculated the CAIs of ORFs for 58 HIV-1 genomic sequences in all collected unstimulated cell types/subtypes/large groups (some cell types like monocytes can be repetitive between bulk and single-cell RNA-seq datasets) according to the previous mentioned bulk RNA-seq datasets and scRNA-seq datasets. As discussed earlier, because HIV-1 can modulate cellular tRNA pools and improve the translational efficiency of its ORFs (Anna et al., 2011), results in this part mainly reflect the capacity of the establishment but not the maintenance of HIV-1 infection. Figure 4 shows the overall CAI distribution of ORFs from all collected HIV-1 sequences in all studied cell types (from either bulk RNA-seq or scRNA-seq datasets). For all analyzed strains, the CAI of a single ORF varies from 0.346 to 0.765. The average CAI among strains and cell types for each ORF fluctuates from 0.500 (vpu) to 0.658 (tat). ORFs tat and rev have the highest CAI, which is consistent with a previous study (Supinya et al., 2018). Compared to the CAIs of endogenous genes exhibited in Figure 2 , in general, HIV-1 ORFs exhibited significantly lower CAI than high-expressed endogenous genes. Results of CAI calculation based on scRNA-seq datasets (see below) are also consistent with this point. As a result, we can postulate that the translational efficiency of ORFs in the HIV-1 genome is lower than most endogenous genes and HIV-1 has been still not well adapted to translation-level regulatory mechanisms of gene expression in unstimulated human cells, at least in cell types we studied. This result may explain why HIV-1 should modulate the cellular tRNA pool to improve its ORFs’ translational efficiency (Anna et al., 2011).

The damage of viruses to cells can be comprehensively measured from two aspects, one is their ability to enter cells, and the other is their destructive ability after entering cells. Thus, we also hoped to use the CAI to indirectly reflect the adaptation of the virus to the translation system of the host cell and the ability of the virus to express proteins and damage host cells after entering. We compared HIV-1 ORFs’ CAIs in several groups of cell types. First, we analyzed HIV-1 ORFs’ CAIs in several types of monocyte-macrophage system cells from bulk RNA-seq data ( Figure 5 ). Among these cell types, the highest CAI appears in Kupffer cells (hepatic macrophages) and I+ dendritic cells. On the other hand, the lowest CAI appears in macrophages differentiated from monocytes in vitro. The second lowest CAI appears in Langerhans cells which are tissue-resident specialized macrophages of the skin. However, it has been reported that in vitro differentiated macrophage is susceptible to HIV-1 and in vitro HIV-1 infection models based on this cell type have been established (Deshiere et al., 2017). Also, during sexual transmission, Langerhans cells are one of the first groups of HIV-1 target cells and their migrations to lymphatic nodes mediate viruses’ infection to CD4+ T cells (Botting et al., 2017). Hence, this result is not consistent with the actual pathogenesis of HIV-1 infection, which provides evidence for the decoupling of CAI and translational efficiency by HIV-1 continuous infection. In addition, according to Figure 5 , we found that the variation of CAI of three ORFs for essential primary protein products (gag, gag-pol and env) is significantly lower than the other six ORFs for non-essential primary protein products (nef, rev, tat, vif, vpr and vpu). This result can be attributed to different mutation rates of genes that encode essential and non-essential protein products in HIV genomes.

The second group we analyzed is a Smart-seq2 based single-cell RNA-seq dataset of the lung (Maynard et al., 2020) which contains 16 types and 3 large groups of cells as mentioned in the ‘Methods’ part ( Figure 6 ). In this dataset, the highest overall CAI of HIV-1 ORFs appears in the group of ‘other T cells’. This group refers to T cells with no detected expression of CD4 and CD8. According to the result of Louvain clustering algorithms performed by Seurat, this group locates in the same group with CD4+ T cells. Hereby, most cells in this group are likely CD4+ T cells whose CD4 expression was not detected for some technical reasons. On the other hand, the CAI of HIV ORFs in all T cells is the second highest and in CD4+ T cells is the third highest among all cell types and large groups. These results are consistent with the conventional opinion that CD4+ T cells are major host cells of HIV-1 in humans. On the other hand, several types of epithelial cells which are not major HIV-1 target cells exhibit relatively low CAIs. However, the CAI of another well-studied target type of cells of HIV-1 (macrophages) is similar to epithelial cells. Finally, it should be noted that different cell types belonging to the hematopoietic lineage exhibit significant variance in HIV-1 ORFs’ CAI. Thus, two cell types originating from the same developmental lineage do not mean that HIV-1 exhibits similar ORFs’ CAI in them.

For another single-cell RNA-seq dataset of PBMC (Deng et al., 2020), postulated susceptibility of cell types based on CAI is also not consistent with previous experiment studies ( Figure 7 ). For HIV-1 ORFs of various strains, their mean CAIs in B cells but not CD4+ or other subtypes of T cells are highest. Also, in two subtypes of monocyte in the blood which have been reported to be host cells of HIV-1, HIV-1 ORFs exhibit the lowest CAI in all cell types. These results imply that CAI is not a good indicator for evaluating cell types’ susceptibility to HIV-1 in peripheral blood.

Then we compared HIV-1 ORFs’ CAIs in four different subtypes of CD4+ T cells in peripheral blood ( Figure 8 ). In this comparison, the overall trend of CAIs for HIV-1 ORFs in these subtypes is Tfh>Treg> CD4+ T nonnaive> CD4+ T naive. The highest HIV-1 ORFs’ CAI in Tfh is consistent with several recent studies which identify Tfh as a highly susceptible subtype for HIV-1 and a major site for HIV-1 replication and production during untreated infection (Perreau et al., 2013). However, it should be noted that the differences in HIV-1 ORFs’ CAI between these subtypes are small (<0.03 for the average CAI of each ORF in different subtypes, corresponding translational efficiency differences less than 1.04 times according to the linear relationship between CAI and translational efficiency of endogenous genes in patient-derived CD4+ T cells from Figure 3H ). In conclusion, these results show that CAI can be used as a reference but is not conclusive evidence to evaluate cells’ susceptibility to HIV-1.

Finally, we focused on two organs that play important roles in metabolic regulation-liver and adipose tissue. We selected four types of nonimmune cells (hepatocytes, CD133+ and CD133- cholangiocytes and hepatic satellite cells) in the liver and adipocytes (the major type of cell that stores lipids in adipose tissue) for CAI calculation and further analysis ( Figure 9 ). Some studies have explored the role of these key metabolic organs in the pathology of AIDS. For adipose tissue, its dysfunction related to HIV-1 infection is common in patients with AIDS (Koethe, 2017) and it has been proposed as a reservoir for HIV-1 (Couturier and Lewis, 2018). Nevertheless, direct infection of HIV-1 to adipocytes has been not reported (Couturier and Lewis, 2018). For the liver, the susceptibility of its components to HIV-1 has been studied. Both in vivo and in vitro studies have shown the ability of HIV-1 to infect hepatocytes (Xiao et al., 2008; Kong et al., 2012; Zerbato et al., 2022). An in vitro study verified that HIV-1 can infect hepatic satellite cells (Tuyama et al., 2010). However, no studies yet have shown that cholangiocytes can be infected by HIV-1. Surprisingly, hepatocytes and adipocytes exhibit high HIV-1 ORFs’ CAI. In fact, hepatocytes exhibit the highest overall CAI in all cell types analyzed in this study. On the other hand, hepatic satellite cells which can be infected by HIV-1 in vitro exhibit relatively lower HIV-1 ORFs’ CAI. This result seems to indicate that the low susceptibility of adipocytes to HIV-1 is due to the lack of receptors rather than unsuited translational regulatory mechanisms.

During the evolution of viruses, their infectious capacity to different target cells changes gradually. These changes may be manifested in the adaptation of the translational regulatory system of host cells. It was reported that patterns of codon usage can reflect basic features of molecular evolution (Sharp and Matassi, 1994). Therefore, it is possible that ORFs’ CAIs of different HIV-1 strains can reflect the phylogenetic relationship between them. For verifying this hypothesis, we performed a hierarchical clustering analysis of HIV-1 strains based on their ORFs’ standardized CAIs in different cell types ( Figure 10 ). However, in the hierarchical clustering analysis, several groups of HIV-1 strains which belong to the same subtype (e.g. L, G, U) are not grouped together. These results indicate that ORFs’ CAIs of HIV-1 strains cannot reflect their phylogenetic history efficiently. This phenomenon may imply that the codon adaptation of HIV-1 ORFs is not selected strongly during the evolution and the adaptation to humans of HIV-1. Thus, it seems efficient translational elongation is not a major limiting factor for HIV-1 infection in humans.

We are now focusing on SARS-CoV-2, the pathogen of the COVID-19 pandemic. We calculated the CAI values of SARS-CoV-2 with five VOCs (Alpha, Beta, Gamma, Delta, Omicron) and a reference (original virus identified in Wuhan) in the lung scRNA-seq dataset (Maynard et al., 2020). As discussed earlier, because SARS-CoV-2 infection does not interfere with the relationship between CAI and translational efficiency, this part may reflect both the establishment and the maintenance of SARS-CoV-2 infection. Figure 11 shows the overall CAI distribution of each ORF of different SARS-CoV-2 subtypes in different lung cells. We can know that, on the whole, the CAI value of the N gene is the highest and the CAI values of E, ORF10 and ORF6 genes are the lowest. However, most of the other ORFs have similar CAIs in the middle. This is basically consistent with the SARS-CoV-2 coding capacity experiment (Finkel et al., 2021), which indicates that the translation efficiency of different ORFs in SARS-CoV-2 is similar. This shows that CAI may be a good indicator of the translation efficiency in SARS-CoV-2. In addition, the overall CAI values of SARS-CoV-2 are lower than most endogenous genes in the previously mentioned three cell types. This result shows the codon adaptation for SARS-CoV-2 in human cells is not sufficient now.

We also found an interesting relationship. Among the essential genes transcribed from subgenome mRNA (sg mRNA), the N gene has the highest expression level, while the E gene has the lowest expression level in the conclusion drawn from the SARS-CoV-2 coding capacity experiment (Finkel et al., 2021). What’s more, the ORF10 and ORF6 genes also show very low expression levels in the coding capacity experiment. Although it is generally believed that CAI does not directly reflect expression levels, these results have an interesting correspondence with CAI results.

In addition, we also counted the relative levels of CAI values corresponding to ORFs of different SARS-CoV-2 subtypes in the lung dataset. The results are shown in Figure 12 . Overall, compared with the reference sequence, the CAI values of other subtypes show a downward trend, which likely indicates that the damage of the virus to the host decreases during the passage. This is consistent with many existing research results (Huang et al., 2021; Barh et al., 2022; McMahan et al., 2022), further proving that CAI can indirectly reflect the damage of the virus. Nevertheless, it should be noted that the mutation of the virus is relatively small now, and its future trend needs further observation. Additionally, in the statistical results, we also find an abnormal situation, that is, the CAI value corresponding to the Omicron ORF7b is significantly higher than that of the reference sequence and other subtypes. The increase in CAI value corresponding to ORF7b may be related to the current pandemics of the Omicron subtype, which needs more experimental research.

Despite significant pulmonary symptoms, COVID-19 is regarded as a systematic disease with symptoms in various organs. For instance, infection of SARS-CoV-2 to the kidney and related acute kidney injury have been reported (Chen et al., 2021). However, renal-related symptoms are less common in COVID-19 than respiratory symptoms. Besides the difficulty of the virus in reaching the kidneys, is there any other explanation for this phenomenon? We analyzed ORFs’ CAI of the original SARS-CoV-2 strain in two cell types from glomerulus (podocytes and mesangial cells; Figure 13 ). Among them, an in vitro study has proved that podocytes can be infected by SARS-CoV-2 with the help of ACE2 and BSG/CD147 at their cell surface (Kalejaiye et al., 2022) but no studies have reported that mesangial cells can be infected by SARS-CoV-2. We found that the ORFs of SARS-CoV-2 have very low CAIs in both two types relative to other cell types. Especially, the lowest overall CAIs of SARS-CoV-2 ORFs appear in podocytes among all analyzed cell types. This fact may indicate the difficulty of SARS-CoV-2 to infect these cell types and help to explain why renal-related symptoms are not as common as respiratory symptoms in COVID-19.

As discussed in the introduction, the inhibition of viral gene translation is a major part of the anti-virus immune response of animal cells (Stern-Ginossar et al., 2019). Therefore, we postulated that the translational efficiency of viral mRNA will be significantly different between uninfected and infected cells. This change may be reflected through CAIs. On the other hand, compared to uninfected cells, CAIs in infected cells can represent the condition in real patients with infectious diseases caused by viruses better. For HIV-1, the correlation between CAI and translational efficiency of endogenous genes is interfered with by viral infection ( Figures 3H and I ), so we did not analyze the change of viral ORFs’ CAIs after HIV-1 infection. However, the correlation between CAI and translational efficiency is maintained in cells infected with SARS-CoV-2. Thus, for studying the effect of viral infection on the translational elongation of SARS-CoV-2 mRNAs, we selected the gene expression datasets of the cells in the control state and the virus-infected state (Daamen et al., 2021; Mulay et al., 2021; Sun et al., 2021; Puray-Chavez et al., 2022), and calculated the CAI values of the SARS-CoV-2 genomes in the corresponding cell datasets. The relative changes of CAI values in the control group and experimental group (virus infection) are consistent in different SARS-CoV-2 subtypes, and the relative positions of CAI values corresponding to ORFs are unchanged. Therefore, using the ORFs of the reference genome for calculation is enough to show the change of cell state after virus infection. It can be seen from Figure 14 that in more cells, the change of CAI is larger than zero (meaning that the CAI of the experimental group is higher than that of the control group), which indicates that maybe after virus infection, the gene expression of cells changes to be more conducive to virus translational elongation. Additionally, in the dataset of HBEC cells (Puray-Chavez et al., 2022), the CAI values basically increase with the infection time and the CAI values in infected cells are all larger than control cells after 48 hours of infection. This is also strong evidence for the above conclusion.

Except for SARS-CoV-2, there are six other coronavirus species that can infect humans and cause disease. Among them, HCoV-229E, HCoV-OC43, HCoV-NL63 and HCoV-HKU1 can only infect human cells and just cause mild diseases, while MERS-CoV, SARS-CoV and SARS-CoV-2 are zoonotic in origin and cause severe respiratory illness and fatalities (Hasöksüz et al., 2020). To investigate whether the CAIs of ORFs of these viruses can reflect their difference in pathogenicity and target cells, we calculated the CAIs of their ORFs in cell types from the lung scRNA-seq dataset according to the procedure mentioned earlier.

In Figure 15 , the CAI statistics of 6 essential genes in different cells are shown by boxplots. Similar to SARS-CoV-2, the overall CAI values of these coronaviruses are lower than most endogenous genes in the three cell types mentioned before. This result implies these viruses have not optimized their codon usage for efficient translational elongation in host cells. From Figure 15 , we can also know that on the whole, the CAI level of MERS-CoV is higher than that of SARS-CoV, and the CAI level of SARS-CoV is higher than that of SARS-CoV-2, which is consistent with the actual pathogenic level of the three (Rabaan et al., 2020). Among the four other human coronaviruses, the overall CAI levels of HCoV-229E and HCoV-OC43 are higher than those of HCoV-NL63 and HCoV-HKU1, which is basically consistent with their actual pathogenic levels (Bouvier et al., 2018). It can be seen that the overall CAI value can reflect the translation level of the virus in the host to a certain extent, and then reflect its pathogenic level. However, we also found that although SARS-CoV-2 can cause pandemics and acute pneumonia, its overall CAI level is slightly lower than HCoV-229E, which means the pandemic and acute pneumonia caused by SARS-CoV-2 may be because of its higher levels of infection and entry ability. But it is also possible that direct comparisons are not appropriate because the viruses infect the host differently (zoonotic/specifically infects humans). In addition, if we compare the relationship among coronaviruses with the same receptor (ACE2: HCoV-NL63, SARS-CoV and SARS-CoV-2; 9-O-acetylsialic acids: HCoV-OC43 and HCoV-HKU1 (Guruprasad, 2021; V'kovski et al., 2021)), the level of CAI value can also perfectly reflect its pathogenic level (overall CAI and pathogenic level: SARS-CoV > SARS-CoV-2 > HCoV-NL63; HCoV-OC43 > HCoV-HKU1).

It can be seen from the previous conclusions that in SARS-CoV-2, the CAI is a suitable way to indirectly measure the translation efficiency. Therefore, we tried to verify if the same conclusion was suitable for other coronaviruses by analyzing the CAI distribution of these coronaviruses in different cell types in the lung. The result is shown in Figure 16 and the data is still derived from 6 shared essential genes. It can be seen that the absolute CAIs of the seven coronaviruses in different cells are somewhat different, but the relative magnitude of the CAI values of each virus for different types of cells is basically stable, which shows the translational adaptation patterns of coronaviruses in human lung cells are likely to be similar.

In addition, we can also find that the CAI values of T cells (lungOtherTcell, lungTcell, lungCD4+T, lungCD8+T), NK cells, and mast cells in lung immune cells are relatively high, suggesting that the destructive power of the coronaviruses inside such cells may be high. This is likely due to the co-evolution of virus and host defense mechanisms during natural evolution. Some existing experimental results can confirm this result, such as SARS-CoV-2, SARS-CoV and MERS-CoV can cause a large decrease in lymphocytes (Gu et al., 2005; Chu et al., 2016; Guan et al., 2020). Infection of host T lymphocytes by SARS-CoV-2 (Shen et al., 2022) and rapid induction of host T lymphocyte apoptosis by MERS-CoV (Chu et al., 2016) have been observed. In SARS-CoV-2 related studies, the damage to host cells is T cell > NK cell > B cell, and for T cell, CD4+ T cell > CD8+ T cell. These results are consistent with the level of their CAI values (Shen et al., 2022). In addition, in the study of MERS-CoV, the virus infection is T cell > NK cell > Monocytes > B cell, which is also consistent with the relative level of CAI values (Chu et al., 2016). Some experiments about SARS-CoV can also support our results, which claim that from their samples, the infected T cells are more than NK cells and B cells (Gu et al., 2005).

However, in the CAI value results, the CAI value of lung AT2 cells is relatively low. And we know that AT2 is one of the main host cells of SARS-CoV-2 and SARS-CoV. This shows that AT2 is their main host cell, probably because the virus can enter such cells more easily. However, its translation efficiency and destruction ability in AT2 cells are probably not higher than those in other cells. In addition, some experiments have shown that T cells are resistant to SARS-CoV infection (Chu et al., 2016). However, in other experiments, SARS-CoV is also found in T cells (Gu et al., 2005). From our results, it is likely that SARS-CoV can adapt to the translation system in T cells better than other cell types. In general, CAI can be used as an auxiliary indicator, together with other indicators, to comprehensively evaluate the ability of the virus to infect and destroy specific cells.

If we evaluate the pathogenic ability of different coronaviruses from the overall CAI value in Figure 16 , we can draw the conclusion similar to Figure 15 . This can also indicate that both overall CAI value and CAI for specific ORFs can reflect the translation level of the virus in the host to a certain extent, and then reflect its pathogenic level.