key: cord-0747468-kkuwcfpy authors: Li, Y.; Lai, D.; Lei, Q.; Xu, Z.; Hou, H.; Shan, H.; Wang, F.; Fan, X.; Tao, S.-c. title: Systematic evaluation of SARS-CoV-2 spike protein derived peptides for diagnosis of COVID-19 patients date: 2020-09-03 journal: nan DOI: 10.1101/2020.09.01.20186387 sha: 74e06c82a78b49e8c71406aa6971d80683f9f41a doc_id: 747468 cord_uid: kkuwcfpy Serological test plays an essential role in monitoring and combating COVID-19 pandemic. Recombinant spike protein (S protein), especially S1 protein is one of the major reagents for serological tests. However, the high cost in production of S protein, and the possible cross-reactivity with other human coronaviruses poses unneglectable challenges. Taking advantage of a peptide microarray of full spike protein coverage, we analyzed 2,434 sera from 858 COVID-19 patients, sera from 63 asymptomatic patients and 610 controls collected from multiple clinical centers. Based on the results of the peptide microarray, we identified several S protein derived 12-mer peptides that have high diagnosis performance. Particularly, for monitoring IgG response, one peptide (aa 1148-1159 or S2-78) has a comparable sensitivity (95.5%, 95% CI 93.7-96.9%) and specificity (96.7%, 95% CI 94.8-98.0%) to that of S1 protein for detection of both COVID-19 patients and asymptomatic infections. Furthermore, the performance of S2-78 IgG for diagnosis was successfully validated by ELISA with an independent sample cohort. By combining S2-78/ S1 with other peptides, a two-step strategy was proposed to ensure both the sensitivity and specificity of S protein based serological assay. The peptide/s identified in this study could be applied independently or in combination with S1 protein for accurate, affordable, and accessible COVID-19 diagnosis. 2 One Sentence Summary: 35 Eight S protein-derived peptides, particularly S2-78 (aa 1148-1159) , are of high 36 performance for diagnosis of COVID-19 as well as discrimination of other coronaviruses. 37 Abstract 38 Serological test plays an essential role in monitoring and combating pandemic. Recombinant spike protein (S protein), especially S1 protein is one of the 40 major reagents for serological tests. However, the high cost in production of S protein, 41 and the possible cross-reactivity with other human coronaviruses poses unneglectable 42 challenges. Taking advantage of a peptide microarray of full spike protein coverage, we 43 analyzed 2,434 sera from 858 COVID-19 patients, sera from 63 asymptomatic patients 44 and 610 controls collected from multiple clinical centers. Based on the results of the 45 peptide microarray, we identified several S protein derived 12-mer peptides that have 46 high diagnosis performance. Particularly, for monitoring IgG response, one peptide (aa 47 1148-1159 or S2-78) has a comparable sensitivity (95.5%, 95% CI 93.7-96.9%) and 48 specificity (96.7%, 95% CI 94.8-98.0%) to that of S1 protein for detection of both 49 COVID-19 patients and asymptomatic infections. Furthermore, the performance of S2-78 50 IgG for diagnosis was successfully validated by ELISA with an independent sample 51 cohort. By combining S2-78/ S1 with other peptides, a two-step strategy was proposed to 52 ensure both the sensitivity and specificity of S protein based serological assay. The 53 peptide/s identified in this study could be applied independently or in combination with 54 S1 protein for accurate, affordable, and accessible COVID-19 diagnosis. 55 56 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 3 COVID-19 is caused by SARS-CoV-2[1,2] and is a global pandemic. By August 7, 2020, 57 18,982,658 cases were diagnosed and 712,266 lives were claimed 58 (https://coronavirus.jhu.edu/map.html) [3] . To put the pandemic under control, one of the 59 essential options is to perform fast, reliable and affordable diagnosis. Although nucleic 60 acid test (NAT) is the reference standard for diagnosing COVID-19 with high sensitivity 61 and accuracy, however, false negative results were commonly observed [4, 5] . The 62 immunological/ serological test, for example, monitoring the SARS-CoV-2 specific IgG 63 and IgM responses, provides important information to improve the accuracy of 64 diagnosis [4, 5] . In addition, serological test is suitable for population screening in high 65 risk regions or among close-contact people, as well as surveillance of the pandemic 66 spreading or assess the infection rate of general population [6] [7] [8] . Moreover, antibody 67 response is reported to associated with disease severity or clinical outcomes [9, 10] . 68 69 S protein is the preferential antigen for serological assay. The key reagent of the S 70 protein based serological assay is the recombinant protein. However, the production of 71 the S protein is difficult and costly [11] . The inconsistency among different manufacturers 72 or even batches might contribute to the variability of commercial assays with the same 73 antigen [7, 12] . Limited production capacity and high cost of recombinant protein 74 preparation is the bottleneck, particularly for remote regions or poor countries. In 75 addition, it should be concerned that the cross-reactivity of infections of other human 76 coronaviruses may cause false positive results, especially for those four common cold 77 causing coronaviruses, i.e., HCoV-OC43, HKU1, NL63 and 229E, which are circulating 78 in population [4, 11, 13] . It was reported that S1, compared with full length S protein, 79 exhibit less cross-reactivity due to the less similarity of S1 subunit among the human 80 coronaviruses than that of S2 [4] . To develop highly specific serological test, more efforts 81 are needed to identify sections of S protein that are highly immunogenic and less 82 homologous to other related coronaviruses [6, 11] . 83 84 Spike protein derived peptides that can elicit antibodies in COVID-19 patients has 85 been reported in several studies [14] [15] [16] , including one of our previous work on epitope 86 mapping with a small sample cohort [17] . For instance, antibody against S2-78 (aa 87 4 1148-1159) and S2-22 (aa 812-823) have high positive rates in COVID-19 patients. 88 However, whether those peptides are suitable for diagnosis is still unknown. Herein, to 89 fully evaluate the diagnostic value of the S protein derived peptides, a total of four 90 cohorts of sera, consisted of 2,434 sera from 858 COVID-19 patients, sera from 63 91 asymptomatic patients and 610 controls were used. Eight peptides were verified to have 92 high potential for diagnosis, particularly, one peptide, S2-78 has a comparable diagnosis 93 performance as that of S1 protein for COVID-19 patients and asymptomatic infections. 94 By combining S2-78 IgG/ S1 IgG with other peptide/s, we purposed a two-step strategy 95 that can ensure both sensitive and specific diagnosis for COVID-19. Organization) for antibody laboratory test [18] . As a result, in total, 729 sera were selected. 108 The control groups include two types of samples. The first type is sera collected from 109 hospitals, including sera from healthy people (n=92), upper respiratory infections (URI, 110 n=104), patients with autoimmune diseases (AID, n=120), lung cancer patients (n=41) 111 and patients with other diseases (n=112) that consist of cardiovascular or cerebrovascular 112 diseases (34.2%), diabetes (9%), non-lung cancers (7.2%) and others. The second type is CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted September 3, 2020. To identify which section of S protein has diagnostic value, it is necessary to survey the 123 entire protein on a systematic way. We took advantage of a previously constructed 124 peptide microarray which has full coverage of S protein [17] , analyzed 55 convalescent 125 sera of COVID-19 patients along with 18 control sera (Cohort 1, Table 1 ). Overall, 126 significant bindings of both IgG and IgM were observed in patient group, while the 127 signals were low in the control group (Fig. 1a) . Several peptides, e. g., S2-78 and S1-93 128 exhibit strong IgG antibody bindings and high response frequencies in patients, we define 129 these peptides as "significant" peptides, these peptides may have diagnostic values (Fig. 130 1b). To test whether the peptide specific IgG antibody responses are concentration 131 dependent, 3x serially diluted peptides (0.9, 0.3 and 0.1 mg/mL) were printed and 132 immobilized on the microarray. As expected, the averaged signals of the patient group, 133 but not the control group, are proportional to the concentrations of the peptides (Fig. 1c) . 134 To extensively evaluate the peptides for diagnostic application, a larger cohort (Cohort 135 2, Table 1 ) of samples were screened by a revised peptide microarray that contains only 136 one peptide concentration (0.3 mg/mL) for high-throughput analysis. To ensure the data 137 generated from different microarrays are comparable, we prepared a positive reference 138 sample by pooling 50 randomly selected patient sera. This reference sample was then 139 tested on all the microarrays for normalization (see methods). Consistent results were 140 achieved for most of the peptides. AUC (area under curve) values of IgG or IgM for each 141 peptide were calculated. Eight peptides, i. e., S2-78, S1-97, S1-93, S1-101, S1-111, S2-97, 142 S1-105 and S2-22 are of high performance, the AUCs of IgG or IgM against these 143 peptides for both Cohort 1 and 2 are above 0.85. Specifically, for Cohort 2 the AUC 144 values with 95% CI (confidential intervals) for S2-78, S1-97, S1-93, S1-101, S1-111, CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . https://doi.org/10.1101/2020.09.01.20186387 doi: medRxiv preprint 6 protein and S1 protein exhibited high performance. We further examined the antibody 150 responses in different groups for each peptide (Fig. 2b-e, Fig. S1 ). Consistently, signals 151 for COVID-19 group are significantly higher than that of the negative samples in all 152 groups. It is noted that the signal intensity for S1 IgG is generally higher than that of any 153 single peptide in the group of COVID-19 patients, this may because there are multiple 154 antibody binding sites on S1 protein. However, slightly higher signal is also observed in 155 control group for protein antigens, demonstrating non-specific binding while it is largely 156 eliminated for synthetic peptides. It is suggested that more sensitive detection platforms 157 or higher antigen concentration might improve the performance of peptides for diagnosis. The diagnostic performance of S2-78 IgG is comparable to that of S1 IgG for COVID-19 160 patients 161 We next focused on S2-78, the peptide of best performance for detection. Optimal 162 Youden index of ROC (receiver operating characteristic) curve was used to set the cutoff 163 value. The specificity, sensitivity and overall accuracy (95% CI) of S2-78 IgG for 164 detection of COVID-19 are 96.7 (94.8-98.0%), 95.5% (93.7-96.9%) and 96% (94.8-97%), 165 respectively, which are slightly lower than that of S1 IgG ( Table 2) . Since serological 166 test is essential for population screening, we calculated the PPV (positive predict value) 167 and NPV (negative predict value) of two assumed prevalence rates. One is 0.04 for 168 general population originated from the situations of Wuhan, China[19] and 169 Netherlands [20] . The other is 0.5 for a high risk population [7] . For prevalence rate of 0.5, 170 both PPV and NPV of S2-78 IgG is similar to that of S1 IgG, however, for prevalence 171 rate of 0.04, the PPV is only 54.7%, although the NPV is extremely high, suggesting the 172 antibody detection of S2-78 could effectively exclude negative ones but may generate 173 high false positive rate at low rate. However, for low prevalence rate, since the number of 174 real positive are very low, it might be acceptable to perform additional test by using other 175 antigens to improve the overall performance. We next investigated the consistency between S1 IgG and S2-78 IgG (Fig. 3a, b) CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . https://doi.org/10.1101/2020.09.01.20186387 doi: medRxiv preprint 7 for S1 IgG but positive for S2-78 IgG, suggesting it will be of diagnostic value to 181 combine S2-78 and S1. It is known that the immune response may correlate to some key 182 clinical parameters, such as gender, disease severity, age and the final outcome [5, 9] . To 183 test whether the positive rate of S2-78 IgG is associated with these clinical parameters, 184 we analyzed the detection sensitivities among subgroups, i. e., male vs. female, ≥60 vs. 185 <60 for age, severe vs. non-severe cases and survivor vs. non-survivor with critical 186 diseases. Similar to that of S1 IgG (Fig. S2a) , no significant difference was observed in 187 all these subgroups. weeks after onset, which is similar as S1 and N IgG (Fig. 3d) . Similar trends were 196 observed for S2-78 IgM (Fig. 3e) and IgG antibodies against other peptides (Fig. S2b) . 197 These observations suggest that the antibodies against S2-78 and other peptides could be 198 applied for monitoring virus specific antibody dynamics. NAT, IgG or IgM[24] . It was shown that for S1 IgG and S2-78 IgG, all samples (n=4) of 209 IgGgroup were negative, while of the 59 IgG + asymptomatic infections, 47 and 45 were 210 positive for S1 IgG and S2-78 IgG, respectively. (Fig 4a, b) . The consistency between S1 211 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . https://doi.org/10.1101/2020.09.01.20186387 doi: medRxiv preprint 8 IgG and S2-78 IgG was also high (93.7%). The contradictory results in IgG + group 212 between our data and the commercial assay may due to the differences of antigens 213 involved. S and N recombinant proteins are used for the commercial assay, for which 214 slightly lower specificity is common [7] . Overall, these results demonstrate the diagnostic 215 and screening value of S2-78 IgG for asymptomatic infections. The diagnostic value of S2-78 IgG was validated by ELISA 218 ELISA (enzyme linked immunosorbent assay) is common for commercial SARS-CoV-2 219 antibody assays [8, 25] . To verify the efficacy and applicability of S2-78 IgG, we 220 established an ELISA assay. Firstly, to test the consistency of the peptide microarray and 221 ELISA, we randomly selected 31 sera from COVID-19 patients of Cohort 1 and tested by 222 ELISA. High consistency was achieved with a Pearson correlation of 0.926 ( Fig. 5a) , 223 demonstrating the validity of the microarray results. To further validate the performance 224 of S2-78 IgG, we screened another independent cohort of samples collected from a 225 different medical center (Cohort 3, Table 1 ). As expected, high performance of S2-78 226 IgG for specific detection of COVID-19 was achieved by ELISA (Fig. 5b, c) . are unique for SARS-CoV-2 is important for diagnosis [11, 13] . To investigate whether the 237 identified peptides are specific to SARS-CoV-2, we performed homology analysis among 238 SARS-CoV-2 and 6 other coronaviruses (Extended data Fig 3) . High homologies were 239 observed for S2-78 and S2-22, suggesting they alone may not suitable for specific 240 detection of SARS-Cov-2 infection. In contrast, other peptides, i. e., S1-93, 241 97,101,105,111 and S2-97, exhibit low similarities with other coronaviruses, particularly 242 the four coronaviruses that cause common cold in human, suggesting they could be 243 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . 9 applied for specifically detection of SARS-Cov-2 infection. S1-97 with the best 244 performance among these peptides was selected for further investigation. Considering the 245 relatively low sensitivity (86.2%) of S1-97 IgG (Fig. 2, Fig. S4 ), we proposed a two-step 246 strategy by combining S1-97 IgG and S2-78 IgG for detection (Fig. 6a) in each group is independent to each other. According to the two-step strategy ( Fig. 6a) , 255 for any given sample, the first step is to detect S2-78 IgG; For the positive ones, the 256 second step is to detect S1-97 IgG. The samples positive for both S2-78 and S1-97 are 257 defined as final positive, while the samples negative for either one is defined as negative. 258 As a result, the sensitivity for SARS-CoV-2 infection detection would be 82.3%, the 259 specificity for related coronavirus infections would be 93.5% (specificity 1), and the 260 specificity for control group would be 99.8% (specificity 2). To further improve the 261 performance of the two-step strategy, a panel (Panel-A) of peptides was composed by 262 bivariate regression analysis based on the specificity of each peptide to SARS-CoV-2. 263 The linear function for Panel-A is y =0.014* x1 + 0.02* x2 -0.003* x3 + 0.006* x4-2.593, 264 where y represents signals of Panel-A and x1, x2, x3, x4 represents the signals of IgG 265 against S1-93, 97, 101 and 105, respectively. The sensitivity and specificity of Panel A, 266 based on the data generated from Cohort 2, is 88.3% and 96.7%, respectively (Fig. 6b, c) . 267 Follow the two-way strategy by combining S2-78 IgG and Panel-A, the final sensitivity 268 for the detection of SARS-CoV-2 infection, specificity for related coronavirus infection 269 (specificity 1), and the specificity for control group (specificity 2) are 84.3%, 96.8% and 270 99.9%, respectively (Fig. 6d) . When S1 protein and Panel-A are combined, the final 271 sensitivity for the detection of SARS-CoV-2 infection, specificity for related coronavirus 272 infection (specificity 1), and the specificity for control group (specificity 2) are 85.3%, 273 96.8% and 99.99%, respectively (Fig. S4b) . These results demonstrate that combination 274 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . https://doi.org/10.1101/2020.09.01.20186387 doi: medRxiv preprint 10 of human coronavirus conserved peptides or S1 protein and other SARS-CoV-2 specific 275 peptides ("significant" peptides) enable specific detection of SARS-CoV-2 infection with 276 a high specificity and acceptable sensitivity. In this study, we took advantage of a peptide microarray of full S protein coverage 280 [17, 26] , analyzed 2,434 sera from 858 COVID-19 patients, sera from 63 asymptomatic 281 patients and 610 controls collected from multiple medical centers. We identified eight 282 12-mer peptides ("significant" peptides) which exhibit high diagnostic values as antigens 283 to detect SARS-CoV-2 specific IgG or IgM. Among the "significant" peptides, S2-78 284 IgG has a comparable diagnosis performance to that of S1 protein for the detection of Recombinant proteins, particular S1, is one of the key reagents to build immunoassay for 292 detecting SARS-CoV-2 IgG, IgM or IgA. However, expressing S1 protein in the right 293 conformation is usually difficult, and in some cases, the antibodies that recognize 294 membrane spike protein are unable to bind recombinant S protein [27] . Moreover, the 295 inconsistency of the proteins from different manufacturers and even different batches 296 from the same manufacturer may result in high variation. High cost and insufficient 297 capacity to produce enough amount of high quality recombinant S1 protein limits the 298 accessibility of the immunoassay in poor or remote regions around the world. 299 Alternatively, peptide-based immunoassay provides a superior choice to S1 protein 300 assays. The reasons are as follows: 1) The peptide synthesis could be easily scaled up 301 when required. A large amount of peptide could be easily synthesized within a very short 302 period of time, if necessary, the peptide could be synthesized in a GMP facility; 2) The 303 consistency and purity of peptide synthesis is high, there is almost no batch-to-batch 304 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . https://doi.org/10.1101/2020.09.01.20186387 doi: medRxiv preprint 11 variation; 3) Peptide is very stable, as reagent, it could be easily stored and transported; 4) 305 The cost of peptide is about 2-3 magnitude lower than that of S1 protein. We identified S2-78 as a candidate of high diagnostic value. Applying large cohorts of 308 COVID-19 patients and a variety types of controls, we comprehensively verified S2-78 309 IgG as a good candidate for diagnosis of COVID-19, with comparable specificity and 310 sensitivity to that of S1 IgG. Expectedly, we found that the overall consistency of S2-78 311 IgG and S1 IgG is high, suggesting S2-78 has the potential to replace S1 protein. It is 312 notable that, the signal intensity level of S2-78 IgG is lower than that of S1 IgG, which is 313 reasonable since there are multiple sites on S1 protein that could be recognized by 314 antibodies in COVID-19 sera. However, because it is hard to ensure extremely high 315 purity when purifying recombinant proteins, the background of S1 IgG for control group 316 is also higher than that of peptides. These results imply the sensitivity of S2-78 IgG might 317 be further evaluated when raise the antigen concentration or adopt a more sensitive 318 platform for detection, such as electrochemical platform or single-molecule detection 319 technologies [28, 29] . Indeed, we also performed S1 based ELISA assay by a commercial 320 kit with the same set of samples, the overall performance of S2-78 IgG based ELISA is 321 comparable to that of the commercial kits (data not shown). is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . 12 336 S protein shares high sequence similarities with other seasonal circulated human 337 coronaviruses. Theoretically, cross-reactivity may exist when S/S1 is applied as antigen 338 for immunological test, thus cause false positive. It is thus necessary to pinpoint specific 339 regions/ sites of SARS-CoV-2 S protein and eliminate the potential cross-reactivity. 340 Selection of peptides with high antibody responses and low sequence similarities with 341 other coronaviruses will improve the performance of diagnosis. Among the identified 342 peptides, except S2-78 and S2-22, other peptides are very distinct to the four circulating 343 human coronaviruses, implying that they could be served as the specific antigen to 344 eliminate the potential cross-activity. However, these peptides exhibit slightly lower 345 sensitivity and specificity. To take advantage of these "significant" peptides, we proposed 346 a two-step strategy that combine S2-78 with other significant peptide/s to discriminate 347 SARS-CoV-2 from related infections as well as non-infections. This study could be 348 further strengthened by testing sera collected from the infections of common human 349 coronaviruses, i. e., HCoV-OC43, HKU1, NL63 and 229E. In summary, we identified and verified eight peptides derived from S protein that 352 exhibit high diagnostic values. These peptides might be used in different circumstances 353 alone or in combination as candidates to build immunoassay/s for monitoring In comparison to the current protein based immunoassays, the peptide based assays will 355 be highly affordable and accessible. The peptide-BSA conjugates as well as S1 protein, RBD protein and N protein of 390 SARS-CoV-2, along with the negative (BSA) and positive controls (anti-Human IgG and 391 IgM antibody), were printed in triplicate on PATH substrate slide (Grace Bio-Labs, 392 Oregon, USA) to generate identical arrays in a 1 x 7 (for the peptide microarray with 393 three concentrations) or 2 x 7 subarray format (for the peptide microarray with one 394 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. Table S1 . Peptides of SARS-CoV-2 spike protein used in this study 480 481 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted September 3, 2020. . Figure 6 . Discrimination of related coronaviruses with combination of peptides. A. Illustration of the strategy with combination of S2-78 IgG and S1-97 IgG based on the results of Cohort 2 (729 patients and 542 controls). For any given sample, the first step is to detect S2-78 IgG response. For the positive ones, the second step is to detect S1-97 IgG response. The blue numbers indicate the positive or negative rates for SARS-CoV-2 infections, while the green numbers indicated the assumed positive or negative rates for samples not belong the SARS-CoV-2 group. The last step summaries the overall sensitivities and specificities for different sample groups. B. Signals of Panel-A (S1-93, 97, 101 and 105 in COVID-19 patients and control groups. C. S1-93_IgG S1-97_IgG S1-101_IgG S1-105_IgG S1-111_IgG Figure S2 . Performance of peptides and S1 for diagnosis in subgroups. A. forest plot of sensitivities of S1 IgG in different subgroups, i.e., age, gender, severity and outcome. The dots indicate the sensitivities while the error bars indicate the 95% CI. The exact values are also provided. P values were calculated with χ2 test. B. Graph of positive rates of IgG antibodies against the indicated peptides versus days after symptom onset in 2,360 serum samples from 784 patients. Figure S4 . Discrimination of related coronaviruses by combining S1 and peptides of Panel-A. A. ROC of S1-97 IgG based on the results of Cohort 2. B. The combination of S1 IgG and Panel-A based on the results of Cohort 2 (729 patients and 542 controls). A pneumonia outbreak associated with a 485 new coronavirus of probable bat origin A new coronavirus associated with human respiratory 488 disease in China An interactive web-based dashboard to track 490 COVID-19 in real time Antibody responses to SARS-CoV-2 in patients 496 with COVID-19 HCoV-OC43 HCoV-HKU1HCoV-NL63 SARS-CoV-2 HCoV-OC43 HCoV-HKU1 HCoV-229E S1 SARS-CoV-2 SARS-CoV HCoV-OC43 HCoV-HKU1 3.0% S1