key: cord-268098-71g1w1mc
authors: Beckman, M. F.; Igba, C. K.; Mougeot, F. B.; Mougeot, J.-L.
title: Comorbidities and Susceptibility to COVID-19: A Generalized Gene Set Meta-Analysis Approach
date: 2020-09-15
journal: nan
DOI: 10.1101/2020.09.14.20192609
sha: 
doc_id: 268098
cord_uid: 71g1w1mc

Background The COVID-19 pandemic has led to over 820,000 deaths for almost 24 million confirmed cases worldwide, as of August 27th, 2020, per WHO report. Risk factors include pre-existing conditions such as cancer, cardiovascular disease, diabetes, obesity, and cancer. There are currently no effective treatments. Our objective was to complete a meta-analysis to identify comorbidity-associated single nucleotide polymorphisms (SNPs), potentially conferring increased susceptibility to SARS-CoV-2 infection using a computational approach. Results SNP datasets were downloaded from publicly available GWAS catalog for 141 of 258 candidate COVID-19 comorbidities. Gene-level SNP analysis was performed to identify significant pathways by using MAGMA program. SNP annotation program was used to analyze MAGMA-identified genes. COVID-19 comorbidities from six disease categories were found to have significant associated pathways, which were validated by Q-Q plots (p<0.05). The top 250 human mRNA gene expressions for SNP-affected pathways, extracted from publicly accessible gene expression profiles, were evaluated for significant pathways. Protein-protein interactions of identified differentially expressed genes, visualized with STRING program, were significant (p<0.05). Gene interaction networks were found to be relevant to SARS and influenza pathogenesis. Conclusion Pathways potentially affected by or affecting SARS-CoV-2 infection were identified in underlying medical conditions likely to confer susceptibility and/or severity to COVID-19. Our findings have implications in COVID-19 treatment development. Keywords: SARS-CoV-2, COVID-19, comorbidity, susceptibility, severity

Visualization of protein-protein interaction networks was completed using STRINGv11.0 [31] program by testing different confidence levels to identify ontologies of biological significance for the significant pathways associated with comorbidities.

Possible comorbidity significant associated gene sets/pathways were checked for quality control by generating Quantile-Quantile (Q-Q) plots using observed quantiles and residual Z-scores of genes within the gene set, based on the MAGMAv1.07b publicly available Rv3.6.2 script (posthoc_qc_107a.r) [32, 33] .

Ensembl's Variant Effect Predictor program (VEP) [34] was used to analyze MAGMAv1.07b annotation files for each gene set associated with comorbidities [35] .

MAGMAv1.07b annotation files were converted into VEP format using a bash script. All converted annotation files were uploaded into VEP online tool separately. VEP summary statistics and analysis tables were downloaded for the comorbidities' associated genes and pathways found significant by MAGMAv1.07b. Corresponding tables were merged [36, 37] containing annotated gene symbols and Entrez gene identifiers for all human genes were used to retrieve missing gene identification [38] . These tabular (.csv) files were merged and loaded into Rv4.0.2. Entrez gene IDs were matched to gene symbols from VEP analysis files to identify Affymetrix gene symbols. Genes and their corresponding Entrez ID's were then matched to significant genes' Entrez IDs found through combined MAGMAv1.07b -STRING analysis.

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org /10.1101 was used to test the top 250 human mRNA gene expressions for each comorbidity based on available human data using NCBI GEO[39] , by only including comorbidities that had significant pathways identified by MAGMAv1.07b and VEP STRING analyses. Human mRNA expression datasets comparing disease group to healthy controls since 2010 were searched. If no datasets were available post-2010 the latest dataset was downloaded using characteristics described for prior datasets. For diseases with no publicly available datasets comparing healthy controls to disease type, the newest, most relevant dataset was used. Tissue types used for analysis included: (i) peripheral blood mononuclear cells (PBMCs), (ii) cancer tissues, (iii) adipose tissue, (iv) pulmonary tissue, (v) post-mortem brain tissue, (vi) cardiovascular tissue, and (vii) blood stem cells.

For each comorbidity, human mRNA gene expression data corresponding to average log-fold change (aLFC) were formatted for clustering of genes identified by MAGMAv1.07b and VEP and subsequently matched to STRING protein-protein interactions. Gene weights were added manually to account for duplicate genes in the dataset. Genes were mean centered and normalized. Hierarchical clustering was completed using a similarity metric of Manhattan cityblock distance for genes and arrays with average linkage via Cluster3.0v1.59. Clusters were visualized using heatmaps created using JavaTreeViewv1.1.6r4 [40] . Clustered groups of genes for MAGMAv1.07b and VEP genes were run separately through GeneCodisv4.0 online tool [41] for identification of possible biological processes or pathways involved in viral infection [42] . Google searches including HUGO gene symbol and either "influenza" or "SARS" [46, 47] . Risk of bias was assessed according to "Cochrane's Handbook for Systematic Reviews of Interventions" [48] . Human tissue expression relevant to COVID-19 for genes with direct involvement was validated using Ensembl Expression Atlas [49, 50] . Genes not generally expressed in central nervous, cardiovascular, or pulmonary systems were removed from the dataset. Visualization of protein-protein interaction network of genes directly involved with influenza and SARS (caused by SARS-CoV-1) was completed using STRINGv11.0 using an interaction score of 0.400 [31] . All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org /10.1101 /10. /2020 

The overall computational analytical design and associated primary results are presented in Figure 1 .

To conduct generalized gene set analysis, we retrieved publicly available GWAS catalog datasets for 141 out of 258 COVID-19 possible comorbidities/underlying medical conditions. The 141 comorbidities were grouped into 8 categories by disease type based on organ most affected (Table S1 ). Following our MAGMA analysis (Figure 1 : Flowchart section A), gene set and Reactome gene level analyses yielded 69 pathways representing 119 significant genes (p<0.05). These pathways were significant for 22 COVID-19 comorbidities representing 6 disease categories, namely, cancer (n=9); cardiovascular (n=4); neurologic/mental (n=3); respiratory (n=2); skin/musculoskeletal (n=1); autoimmune/endocrine/metabolic (n=3). Reactome significant pathways and genes obtained through MAGMAv1.07b gene-level analysis from Enrichment Map are shown in Tables 1a and b. Using STRINGv11.0 program with the highest confidence interaction score (CIS) of 0.9, processing of the 119 genes yielded a protein-protein interaction network of 70 genes, which was found to be highly significant based on hypergeometric test with Benjamini-Hochberg correction (p=4.36x10 -11 ) (Figure 2a) . The top Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway, identified by using STRINGv11.0, corresponded to Epstein-Barr virus infection with a false discovery rate of 6.72x10 -9 .

Verification of significant pathways using Q-Q plots showed a high association between genes and their relative gene ontology defined pathways, since all plots show a distribution of residual z-scores deviating from the diagonal early on. There were no Q-Q plots with any ambiguous feature. Significant genes had high levels of association with each pathway. Q-Q plots of more than five genes, representing the pathways ontologies "post-translational protein modification", All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org /10.1101 /10. /2020 "translocation of ZAP-70 to immunological synapse", "metabolism" and "cell cycle" and associated possible COVID-19 comorbidities (including asthma), are described in Figure S1 .

Annotation files were converted for 134 of the 141 comorbidities with GWAS catalog datasets available (Figure 1 : Flowchart section B). Of 3704 HUGO gene symbols extracted from VEP, 2996 corresponding Entrez gene IDs were identified using Affymetrix human genome annotation file. Of these gene IDs, 50 were matched with the 119 significant genes identified by MAGMAv1.07b for the 22 comorbidities with significant pathways ( Table 2) . Of the 50 genes, all were included in a protein-protein interaction network of 55 genes using a low CIS in STRINGv11.0 (Figure 2b) . The top KEGG pathway identified using STRINGv11.0 was HTLV-1 infection with a false discovery rate of 4.38x10 -7 using hypergeometric test with Benjamini-Hochberg correction.

GEO human mRNA expression datasets were retrieved for 19 of 22 comorbidities. A description of GEO datasets is presented in Table S2 . Using 119 MAGMAv1.07b identified genes ( Figure   1 : Flowchart section A), JavaTreeViewv1.1.6r4 clustered 4 of 9 cancer types in the heatmap (partial view in Figure 3a , full heatmap in Supplementary Image). Also, interstitial lung disease, multiple sclerosis, asthma, obesity, and heart failure were clustered (Figure 3a) . VEP STRING matched genes (n=50) also clustered 4 of 9 cancer types and clustered interstitial lung disease, multiple sclerosis, asthma, obesity, and heart failure together (Figure 3b ). In both heatmaps Nucleoporin 160 (NUP160), Nucleoporin 153 (NUP153), Fibroblast Growth Factor [51] [52] [53] [54] . We also identified three genes KPNB1, Signal Transducer and Activator of Transcription 3 (STAT3), and Interleukin 2 Receptor Subunit Alpha (IL2RA) shown to play a significant role in SARS [55] [56] [57] . Genes identified as being possibly directly associated with influenza and/or SARS are shown in Table S3 . STRING protein-protein interaction network yielded 38/46 (82.6%) genes involved in influenza and 15/17 (88.2%) genes involved in SARS, using an interaction score of 0.4 (Figures 4a and b) . No GWAS study was found for SARS-CoV-1 infection to identify possible susceptibility genes within the 119 genes. Additionally, no studies were found to be at high risk for bias (Table S4) .

This is the first study conducting generalized gene set analysis on a broad spectrum of possible COVID-19 comorbidities, with the prospect of identifying comorbidity-specific genes that could impact infection by SARS-Cov-2.

Starting with a list of 258 diseases, our MAGMA pipeline was able to identify 69 significant Reactome pathways with a total of 119 significant genes corresponding to 22 comorbidities that might have implications in predicting the severity of SARS-CoV-2 infection (Figure 1 , Table 1,   Table S1 ). Of the 22 comorbidities, we were able to validate pathways associated with cardiovascular disease, diabetes, obesity, and pulmonary diseases. Cardiovascular diseases identified included heart failure, atherosclerosis, Kawasaki's disease, and hypertension.

Pulmonary diseases included asthma and interstitial lung disease. Cancer has been reported as a possible risk factor for COVID-19 [9] . We were able to identify nine cancers with GWAS data and significant associated pathways including acute myeloid leukemia, renal cell cancer, small cell lung cancer, and lung cancer. Furthermore, the known COVID-19 comorbidities, hypertension, obesity and diabetes had significant pathways and genes.

While Q-Q plots indicated validity of our findings, caution for interpretation of Q-Q plots must be used as these plots are normally used for pathways containing many genes. To a certain degree, these allow us to convey a certain level of confidence that there is a true association between gene and pathway [33] . In our analysis, however, less genes identified allowed us to narrow possible gene targets and pathways. Indeed, certain genes identified in our study may have significant biological relevance to infection by SARS-COV-2. For instance, sialyl All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . 1 0 transferase ST6 N-acetylgalactosaminide alpha-2,6-sialyltransferase 3 (ST6GALNAC3) was found significant in the post-translational protein modification pathway ( Figure S1 ). Another sialyl transferase, ST6GALNAC1, has been previously investigated as a drug target against infection of smooth airways epithelial cells by influenza virus [58] . It remains, however, to be determined whether ST6GALNAC3, generally expressed at high levels in renal cell cancer[59], plays a significant role in COVID-19 pathogenesis. Interestingly, aLFC gene expression of ST6GALNAC3 was positive in our analysis for 8 of 19 comorbidities (including renal cell cancer), namely, in tumor, pulmonary, brain, adipose tissues and PBMCs (Figure 3a&b ).

STRINGv11.0 analysis produced significant enrichment for both MAGMAv1.07b genes and VEP matched genes containing SNPs that had characteristics of deleterious effects ( Table 2) .

Therefore, we believe the interactions among genes from significant pathways from MAGMA and matched VEP genes are likely not due to chance and that these genes are biologically connected. Furthermore, STRINGv11.0 analyses identified top KEGG pathways including, Epstein-Barr virus pathway (MAGMA genes), and HTLV-1 pathway (VEP matched genes).

STRING was able to cluster 70 genes into four functional groups among the 119 MAGMA significant genes: cell regulation and immune response, cell transport and nervous tissue function, protein homeostasis and gene expression, transcriptional regulation and RNAmediated silencing (Figure 2a) . Additionally, NUP160, NUP153, and KPNB1 clustered tightly together in the cell transport and nervous tissue function group.

STRINGv11.0 analysis of the 50 VEP matched genes with a lower confidence interval of 0.150 was required to obtain sufficient network connections for interpretation. Although network analysis may be subjective and is dependent on established knowledge, it is important to note that the enriched protein-protein interaction p-value was statistically significant. For the VEP matched gene STRINGv11.0 analysis, there were four distinct biological groupings recognized within the mapped network based on the closeness of protein interactions (Figure 2b ). Those groupings were (i) antigen specific immune response, (ii) cell division and molecule formation/development, (iii) cell growth, survival, proliferation, motility, and morphology, (iv) and voltage gated ion channel transmembrane proteins. Notably, one of the comorbidities with significant associated pathways, breast cancer, contained SNPs affecting Solute Carrier Family 4 Member 7 (SLC4A7) and Solute Carrier Family 24 Member 3 (SLC24A3) genes. These genes are involved with sodium, calcium, and potassium ion transport and play a role in the malignant progression of breast cancer [60] . In addition, Euchromatic Histone Lysine Methyltransferase 2 All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . Epithelial signaling by fibroblast growth factors is required for effective recovery from lung injuries resulting from influenza infection [51] . Our analysis coincides with previous findings linking induced inactivation of FGFR2 with increased mortality and influenza-induced lung injury [51] . KPNB1 (Figure 2a) . Furthermore, KPNB1 is involved in the early stage of influenza virus replication via nuclear trafficking, by way of, nuclear import of viral cDNA or viral/host proteins into the host chromosome [52, 53] .

Based on previous studies, the interaction between NUP153 and KPNB1 has been investigated in relation to nuclear transport [63] . The degradation of NUP153 in influenza virus A infected cells, such as Madin-Darby canine kidney II and human lung epithelial cells, results in an enlargement and widening of nuclear pores [54] . This disease process allows viral ribonucleoprotein complexes to be exported from the nucleus to the plasma membrane [54] .

Additionally, NUP160 has been shown to work in conjunction with NUP153 to mediate nuclear import and export [64] . Therefore, degradation of one or both can prevent the import of signal transducers and activators of transcription, reducing effectiveness of the anti-viral interferon All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org/10.1101/2020.09.14.20192609 doi: medRxiv preprint response [65] . Our results support the interactions between these genes and viral respiratory 1 3

Further research is needed to confirm these genes (or associated regulations) as possible drug targets for SARS-CoV-2 infection.

While there is no shortage of publicly available data, not all diseases have the same level of dedicated research. Therefore, not all possible comorbidities had publicly available SNP datasets from GWAS catalog or human mRNA gene expression datasets from NCBI's GEO datasets database. This resulted in a large decrease from 258 possible comorbidities to 141. Additionally, we were only able to use 19 of 22 significant comorbidities for GEO2R analysis and heatmap visualization. Another caveat is that GEO2R mRNA expression datasets have been generated through different independent studies using different genomic platforms and analysis pipelines, so that optimal normalization of raw data cannot be implemented. Little is still known about COVID-19 pathogenesis, although research on the matter has increased greatly since the beginning of the pandemic. 

All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org /10.1101 /10. /2020 Dr. Jean-Luc Mougeot and Dr. Farah Bahrani Mougeot conceived the study, contributed to the design of the analytical strategy, data interpretation and verification. Micaela Beckman designed the overall analytical strategy, conducted most computational analyses and data interpretation, and wrote the manuscript draft. Chica Igba contributed to the design of analytical strategy, conducted analyses and participated into data interpretation. All authors had significant contributions in writing and revisions of main manuscript, tables and figures. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org /10.1101 /10. /2020 preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org /10.1101 /10. /2020 176814; 174048; 176409; 174143; 179419; 174048; 113507; 5687128; 176412; 176408; 453276; 425407; 425393 Activation of APC C and APC C: Cdc20 mediated degradation of mitotic proteins; Cyclin B; mitotic proteins; cell cycle proteins; cell cycle protein prior to satisfaction of cell cycle checkpoint; Phospho-APC C mediated degradation of Cyclin A; Phosphorylation and regulation of APC C between G1 S and early anaphase; E2F enabled inhibition of prereplication complex formation; MAPK MAPK4 signaling; Regulation of mitotic cell cycle; SLC-mediated transmembrane transport; Transport of inorganic cations anions and amino acids oligopeptides 3.57E-11; 3.32E-5 422475; 204998; 73887; 1266738; 416482; 9675108; 193648; 193704; 194840; 194315 Axon guidance; Cell death signaling via NRAGE, NRIF, and NADE; Death receptor signaling; Developmental biology; G alpha (12 13) All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020 . . https://doi.org/10.1101 preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org/10.1101/2020.09.14.20192609 doi: medRxiv preprint 

A list of candidate comorbidities (n=258) possibly associated with increased severity/infectivity of COVID-19 were curated. SNPs associated with comorbidities with available GWAS catalogdata (n=141) were analyzed. Multi-marker Analysis of GenoMic Annotation (MAGMA) was performed. SNPs were annotated to genes using NCBI gene reference file (NCBI37.3.gene.loc). In MAGMAv1.07b, gene set/pathway analysis was performed for which each SNP that was identified, using the "multi-mean=snp-wise" model-generated results, taking into account ethnicities associated with a possible comorbidity. Gene-level analysis was completed using Reactome pathways retrieved from Enrichment Map program. STRINGv11.0 protein-protein interaction program was used to visualize the network of 119 significant genes.

Quantile-quantile (Q-Q) plots in Rv3.4.2 for 69 significant pathways were used for quality control. NCBI-gene expression omnibus (GEO) human mRNA differential expression datasets were downloaded via GEO2R for each comorbidity with associated genes/ pathways (n=19 of 22). Human mRNA expression was visualized with a heatmap of the 119 significant genes using Cluster3.0v1.59 and JavaTreeViewv1.1.6r4. Tissue expression relevance to SARS and influenza was determined using DisGeNETv6 and Ensembl Expression Atlas databases.

MAGMAv1.07b annotation files were converted for Ensembl Variant Effect Predictor (VEP) format (n=134 of the 141 GWAS datasets). Gene symbols (n=3704) were extracted for VEP analysis from 22 significant comorbidity-associated genes/pathways per MAGMAv1.07b analysis. Entrez gene IDs (n=2996) were matched to gene symbols using Affymetrix gene symbols annotation files (HG-U133A/B Human Genome Files). STRINGv1.0 protein-protein interaction program was used to visualize the network (n=50 genes).

NCBI-Gene expression omnibus (GEO) human mRNA differential expression datasets were downloaded via GEO2R for each of the 22 comorbidities with associated genes/pathways (n=19 of 22). VEP genes were matched to network genes and formatted for Cluster3.0v1.59. Human mRNA expression was visualized with a heatmap using average log-fold changes (aLFC) in JavaTreeViewv1.1.6r4 (i.e., 50 VEP identified genes matched to 119 MAGMA identified genes). All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org/10.1101/2020.09.14.20192609 doi: medRxiv preprint All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . Genes were filtered by removing those with less than 60% values present, mean centered and normalized. Hierarchical clustering was completed using weights for duplicate/synonymous genes with a similarity metric of city-block (Manhattan) distance for genes and arrays with average linkage through Cluster3.0v1.59 software. Heatmaps were created using JavaTreeViewv1.1.6r4. Yellow depicts positive aLFC, blue depicts negative aLFC, black depicts missing data and values of zero. STRING protein-protein interactions of MAGMAv1.07b identified genes with direct involvement with (a) influenza (n=46) and/ or (b) SARS (n=17) are shown. Level of stringency in STRINGv11.0 program was set to a medium confidence interaction score (CIS) of 0.4 in both influenza and SARS related molecular networks (a, b), resulting in a cluster of 38/46 (82.6%) and 15/17 (88.2%) genes, respectively. All rights reserved. No reuse allowed without permission. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted September 15, 2020. . https://doi.org /10.1101 /10. /2020 

Clinical Characteristics of Coronavirus Disease 2019 in China

World Health Organization: Q&A: Influenza and COVID-19 -Similarities and Differences. www.who.int/emergencies/diseases/novel-coronavirus-2019/question-and-answershub/q-a-detail/q-a-similarities-and-differences-covid-19-and-influenza#:~:text=Mortality for COVID-19,quality of health care

The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application

Review of the 2019 novel coronavirus (SARS-CoV-2) based on current evidence

Evaluation, and Treatment of Coronavirus (COVID-19)

Coronaviruses pathogenesis, comorbidities and multi-organ damage -A review

Prevalence of comorbidities and its effects in patients infected with SARS-CoV-2: a systematic review and meta-analysis

Determinants of Severity in Cancer Patients with COVID-19 Illness

Effect of Underlying Comorbidities on the Infection and Severity of COVID-19 in Korea: a Nationwide Case-Control Study

Understanding modernday vaccines: what you need to know

Vaccine Effectiveness: How Well Do the Flu Vaccines Work

Accessed 1

Identification of information flow-modulating drug targets: a novel bridging paradigm for drug discovery

Network-based relating pharmacological and genomic spaces for drug target identification

Transcriptome sequencing assisted discovery and computational analysis of novel SNPs associated with flowering in Raphanus sativus in-bred lines for marker-assisted backcross breeding

The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease

A survey on the computational approaches to identify drug targets in the postgenomic era

A review on computational systems biology of pathogen-host interactions

Immunoinformatics--the new kid in town

Influenza Research Database: An integrated bioinformatics resource for influenza virus research

Cochrane Handbook for Systematic Reviews of Interventions version 6

Expression Atlas: gene and protein expression across multiple studies and organisms

FGFR2 Is Required for AEC2

Homeostasis and Survival after Bleomycin-induced Lung Injury

Human host factors required for influenza virus replication

A host of factors regulating influenza virus replication

MS Asthma 5581; 3575; 3117;3123; 6891; 3118

Heart failure 64759 TNS3 192154334 IV; DGV Hypertension 776

Lung cancer 79888

We thank Kathleen Sullivan for her editorial expertise in preparation of this manuscript. This work was supported by the Atrium Health Foundation Research Fund (internal).

 DCUN1D5  AGBL1  INPP5F  STAT5A  AGTRAP  PSMC3  CD44  GBA  CTSC  RFWD2  PPP1CB  STAT3  CD86  BTRC  KCNJ16  MTMR2  SHTN1  ACSL3  AGTPBP1  ELOVL7  NCOA1  CACNB2  SLC44A1  AQP9  SYNE1  NR3C2  PPKAR2B  IQGAP2  MUC2  MAPK10  TRIM36  SLC24A3  ST6GALNAC3  CNTRL  IL2RA  ATXN3 -DQA1  IL7R  SH2B1  ADATMS9  ACOXL  LPCAT1  MYL2  OSBPL10  CACNA1D  MCM8  EHMT2  PRKCE  KIF21B  TAP2  EXT1  NCOA1  IQGAP2  MUC2  ADCY7  FGFR2  NGEF  VAC14  TOLLIP  HLA-DRB1  PIK3R2  SLC4A7  NTM  KCND3  MAD1L1  PRKCA  SLC22A1  RARB  VPS45  TNS3  AGTPBP1  NUP160  LAMC1  NUP153  STAT3  IL2RA  CACNB2  SLC24A3  MAPK10  NR3C2  SYNE1