Massive and rapid COVID-19 testing is feasible by extraction-free SARS-CoV-2 RT-qPCR ==================================================================================== * Ioanna Smyrlaki * Martin Ekman * Antonio Lentini * Martin Vondracek * Natali Papanicolaou * Johan Aarum * Hamzah Safari * Shaman Muradrasoli * Jan Albert * Björn Högberg * Björn Reinius ## ABSTRACT Coronavirus disease 2019 (COVID-19) is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The most widely used method of COVID-19 diagnostics is a reverse transcription quantitative polymerase chain reaction (RT-qPCR) assay, to detect the presence of SARS-CoV-2 RNA in patient samples, typically from nasopharyngeal swabs. RNA extraction is a major bottleneck in current COVID-19 testing, in terms of turnaround, logistics, component availability and cost, which delays or completely precludes COVID-19 diagnostics in many settings. Efforts to simplify the current methods are critical, as increased diagnostic availability and efficiency would benefit patient care and infection control. Here, we describe methods to circumvent RNA extraction in COVID-19 testing by performing RT-qPCR directly on heat-inactivated subject samples as well as samples lysed with readily available detergents. Our data, including benchmarking with 597 clinically diagnosed patient samples against a standardised and sensitive diagnostic system, show that direct RT-qPCR is a viable option to extraction-based COVID-19 diagnostics. Significant savings in terms of time and cost can be achieved by embracing RNA-extraction-free protocols, that feeds directly into the established PCR-based testing pipeline. This could aid the expansion of COVID-19 testing. ## INTRODUCTION The recent emergence of the novel human coronavirus in late 2019 in the Wuhan region in China rapidly evolved into a global pandemic, confirmed to have infected nearly 4 million people worldwide (as of May 5th, 2020). The high transmission rates and high proportion of asymptomatic infections led to a massive, worldwide need for rapid, affordable and efficient diagnostic tests, that can be performed in clinical and non-clinical settings [1,2]. Currently, the widely used method of SARS-CoV-2 detection in clinical diagnostics is an RT-qPCR assay, detecting the presence of viral RNA in patient samples. Although RT-qPCR is widely implemented for the detection of pathogens, including viruses [3] in clinical samples, the implementation of the specific assay for the detection of SARS-CoV-2 has only recently been established. The currently used protocol was developed and optimized for the detection of the novel coronavirus at the Charité University Hospital, in collaboration with several other laboratories in Germany, the Netherlands, China, France, UK and Belgium [4]. Additionally, the existing protocol was further optimized by the Center for Disease Control (CDC) in the United States through the comprehensive comparison and validation of alternative available kits for nucleic acid extraction and the use of alternative probe and primer sets for efficient SARS-CoV-2 detection in clinical samples [5–6]. Routinely, the application of qPCR for the relative quantification of a transcript of interest in a sample, is preceded by *(1)* the isolation of total RNA from the sample and *(2)* the use of purified RNA in a reverse-transcription (RT) reaction resulting in complementary DNA (cDNA) from the template RNA, which is then utilized for the qPCR reaction. However, nucleic acid purification and RT of the resulting RNA into cDNA are not only laborious and time-consuming, but the additional steps requiring manual handling can result in experimental errors. In the case of clinical sampling and diagnostics, the use of a single-reaction kit combining the RT and qPCR reactions is therefore customary. Although single-step RT-qPCR removes the need for a separate RT reaction, RNA isolation from clinical samples constitutes a major bottleneck in the diagnostic process, as it remains both manually laborious and expensive. Specifically, both the Charité University Hospital and the CDC protocols require the use of RNA purification kits, which not only results in a significant cost increase but has already led to a major supply shortage of such kits. It is therefore crucial that a new test is not only affordable, quick and efficient, but also that it keeps the use of industrial kits to the minimum. Recent attempts have been made to circumvent RNA extraction in COVID-19 detection [7–9]. Here, we established routines for SARS-CoV-2 RNA-extraction-free single-reaction RT-qPCR testing (**Fig. 1**) on heat-inactivated nasopharyngeal swab samples in transport medium and compared the results with clinically diagnosed patient samples, demonstrating the viability of extraction-free SARS-CoV-2 diagnostics. In addition, we provide data showing that SARS-CoV-2 RT-qPCR can be performed in presence of high concentration of detergent, allowing testing directly on sample lysates. Importantly, our method builds on clinically established protocols and could easily be integrated to expand ongoing testing pipelines. It is also modular and can be incorporated into alternative approaches of detection utilising PCR. ![Figure 1.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/12/2020.04.17.20067348/F1.medium.gif) [Figure 1.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/F1) Figure 1. Schematic overview of SARS-CoV-2 RT-qPCR testing procedure The currently widely used procedure for COVID-19 testing involves: (**a**) Collection of patient material and deposition of potential SARS-CoV-2 viral particles in transport medium. (**b**) Inactivation of the virus by detergent/chaotropic reagents or by heating. (**c**) RNA extraction. (**d-e**) Transfer to PCR-plate (96/384-well) format in which cDNA synthesis by RT and detection by qPCR may take place. Alternatively, detection can be made by sample barcoding and high-throughput DNA sequencing. (**f-g**) Unlike the widely used approach, which includes an RNA extraction step (c) using industrial RNA extraction kits, direct sample testing circumvents this process, by omitting extraction. Instead, after clinical samples are deposited in transport medium, viral particles are inactivated either through heating or by direct lysis in detergent-containing buffer. The inactivated samples are then used for the downstream RT-qPCR diagnostic reaction. ## RESULTS We started by investigating how transport media used for swab collection affect RT-qPCR. To do this, we spiked synthetic full-genome SARS-CoV-2 RNA (SKU102024-MN908947.3, Twist Biosciences) into dilution series of three different transport media (Virocult MED-MW951S, Sigma; Transwab MW176S, Sigma, and Eswab 482C, COPAN) used for clinical sampling at the time of the study (Karolinska University Hospital, Stockholm, Sweden). Each mock sample contained 50,000 synthetic SARS-CoV-2 gRNA copies and 95-0.1% medium, corresponding to 47.5-0.05% medium in the RT-qPCR reaction. We performed single-reaction RT-qPCR using 10µl sample in a 20µl reaction and the CDC nucleocapsid 1 (N1) primer-probe set (**Table 1**, Methods) and recorded cycle threshold (CT) values for the dilution series of the media. We observed inhibitory effects in all three media and, importantly, pronounced variation between media (**Fig. 2a**). Virocult and Transwab demonstrated similar profiles of inhibition, resulting in +2-3 CT at the highest medium concentrations and minimal inhibition at concentrations below 30% medium in the RT-qPCR reaction. Eswab completely impeded detection at high concentrations but reached a similarly low level of inhibition as Virocult and Transwab at 25% concentration in the reaction. For RT experiments relying on synthetic RNA, it is vital to exclude the possibility of lingering DNA template. Therefore, we additionally performed RT and qPCR reactions in two separate steps (Methods) including RT+/– controls, which demonstrated the lack of DNA amplification signal when the reverse transcriptase was excluded from the reaction (**Supplementary fig. 1a**). Together, these results indicate minimal or no inhibition of Virocult, Transwab and Eswab at ≤25% in the RT-qPCR reaction, corresponding to ≤5µl sample in a 20µl SARS-CoV-2 RT-qPCR reaction. ![Supplementary figure 1.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/12/2020.04.17.20067348/F6.medium.gif) [Supplementary figure 1.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/F6) Supplementary figure 1. (**a**) Amplification plots showing normalized reporter value (ΔRn, linear scale) as a function of qPCR cycle for two-step RT and qPCR performed on synthetic full-genome SARS-CoV-2 RNA (SKU102024-MN908947.3, Twist Biosciences) including (RT+) or excluding (RT-) the reverse transcriptase. (**b**) Agarose gel electrophoresis of RT-qPCR products for reactions described in Fig.2e, but in this case using primers and probes for gene E. (**c**) Amplification plots showing normalized reporter value (ΔRn, linear scale) as a function of qPCR cycle for the samples described in Fig. 2e-f, but using two-step RT and qPCR instead of the single-step TaqPath RT-qPCR (Methods). (**d**) Relative cycle threshold (CT) values of SARS-CoV-2 RT-qPCR using the N1 primer-probe set and a dilution series of the chemical RNase inhibitor polyvinylsulfonic acid (PVSA). To test the inhibition of PVSA in RT and PCR reactions separately, we used either synthetic full-genome SARS-CoV-2 RNA (blue dots and line line) or plasmid DNA (CDC positive control) (blue dots and line) as templatein the RT-qPCR reaction. CT values are shown relative to reactions without PVSA (conc.= 0). The lines indicate the median of triplicate measurements. View this table: [Table 1.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/T1) Table 1. Primers and probes used for SARS-CoV-2 RT-qPCR. ![Figure 2.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/12/2020.04.17.20067348/F2.medium.gif) [Figure 2.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/F2) Figure 2. SARS-CoV-2 hid-RT-qPCR on nasopharyngeal swab samples (**a**) CT values from RT-qPCR performed on dilution series of transport medium (Virocult, Transwab and Eswab) using 50,000 spiked copies of synthetic full-genome SARS-CoV-2 RNA and the N1 primer-probe set. Lines represent the mean of duplicates, shown individually as dots. ND: not detected. (**b**) Bar plots of CT from SARS-CoV-2 RT-qPCR on clinical nasopharyngeal swabs inactivated with MagNA Pure 96 External Lysis Buffer (ELB) or heat (65°C 30 min). Dots indicate CT of duplicates and crosses indicate CT values from diagnostics performed on extracted RNA. Positive controls were extracted RNA from a positive sample (P) and a CDC positive control DNA plasmid (CDC+). Negative controls were extracted RNA from a negative sample (N) and water (H2O). ND: not detected. (**c**) Amplification plots showing normalized reporter value (ΔRn, linear scale) as a function of qPCR cycle for the experiment and samples described in (b). (**d**) Bar plots of CT values of 11 positive nasopharyngeal swab samples using primer-probe sets targeting SARS-CoV-2 gene E, N, and RdRP. (**e**) Boxplots of CT difference in same samples comparing E and N1 with the RdRP primer-probe-set. (**f-g**) Line charts of CT from individual clinical samples (coloured lines) using variable amount of sample input. Shown as absolute CT (f) or CT relative to the 10μl input (g). (**h**) Scatter plots of CT values from clinical diagnostics performed on extracted RNA (y-axis) and hid-RT-qPCR (x-axis) of 85 nasopharyngeal swab samples, shown for different primer-probe set comparisons. Rho indicates Spearman correlation of positive samples. ND: not detected. (**i**) Heatmap of CT values from diagnostics performed on 85 clinical samples using extracted RNA (E, RdRP) and hid-RT-qPCR (N1, RdRP), ranked by E gene CT. Control for sample integrity by RT-qPCR for RNase P in the same samples shown on the right. Two patients marked with asterisk (*) were negative in extractionbased diagnostics but positive by hid-RT-qPCR but later re-tested and confirmed to be SARS-CoV-2 positive (see Results). The patient marked with a ring (o) was not re-tested. Three samples marked with hash (#) were called COVID-19 positive by routine diagnostics but not by any primer-set in hid-RT-qPCR. (**j**) Scatter plot of CT values from 19 matched fresh (y-axis) and freeze-thawed (x-axis) extracted samples, using the E gene (cross) and RdRP (star) primer-probe sets. ND: not detected. hid-RT-qPCR shown in this figure was performed on previously diagnosed frozen samples. View this table: [Table 2.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/T2) Table 2. Heat-inactivation conditions tested for SARS-CoV-2 hid-RT-qPCR optimisation. Related to Fig. 3b-e, with corresponding colour code to the left in the table. To test whether direct RT-qPCR could accurately detect the presence (or absence) of SARS-CoV-2 in clinical samples we started by obtaining five aliquots of nasopharyngeal swab samples stored in transport medium at –20°C. Aliquots of the same samples had previously been clinically diagnosed using conventional RNA extraction (MagNA Pure 96 DNA and Viral NA SV Kit, Roche Diagnostics 06543588001) followed by RT-qPCR, calling three patient samples as SARS-CoV-2 positive and two as negative to the virus (Clinical diagnostics performed at the Karolinska University Hospital, Stockholm). We inactivated the nasopharyngeal samples either by adding MagNA Pure 96 External Lysis Buffer (Roche, 06374913001), used in conventional RNA purification, or by heating at 65°C for 30 min, and performed RT-qPCR using 3µl of sample. We observed a lack of amplification in SARS-CoV-2 positive samples inactivated with External Lysis Buffer (**Fig. 2b-c** and **Supplementary fig. 1b**). However, RT-qPCR performed directly on heat-inactivated samples correctly detected SARS-CoV-2 in all positive samples and lacked signal in the negative samples and controls (**Fig. 2b-c**). This indicated the viability in further exploring heat-inactivated direct RT-qPCR (hid-RT-qPCR) as a method to detect SARS-CoV-2 in clinical samples. We also tested two-step RT and qPCR for SARSCoV-2 detection on the same clinical samples (Methods), correctly detecting the viral presence and absence (**Supplementary fig. 1c**). Next, we tested primer-probe set performance in hid-RT-qPCR using nasopharyngeal swab samples and primers-probe sets targeting the SARS-CoV-2 genes RNA-dependent RNA polymerase (RdRP), envelope (E), and nucleocapsid (N1) (**Table 1**). We obtained additional heat-inactivated (65°C 30min) nasopharyngeal swab samples called as SARSCoV-2 positive in previous clinical diagnostics (Methods). We observed, in our setting, a modest difference between N1 and RdRP in hid-RT-qPCR (mean and median CT difference to N1: 0.63 and 0.27, *P*=0.032, Wilcoxon signed-rank test) while the E-gene set appeared at considerably higher CT values than the other primer-probe sets (mean and median CT difference to N1: 2.9 and 1.7, *P*=0.00098, Wilcoxon signed-rank test) (**Fig. 2d-e**). Lower performance of primers and probes targeting the E gene is in line with previous results [10]. We argue that using short amplicon targets is critical for hid-RT-qPCR due to the expected RNA fragmentation during heating, while considerations on the amplicon length should be less important for PCR amplification performed on extracted RNA from fresh samples. Due to the superior performance of N1 in hid-RT-qPCR, we focused on this primer-probe set in the deepened analyses of hid-RT-qPCR. Our dilution experiments of medium and spike-in synthetic SARS-CoV-2 RNA had shown limited inhibition at ≤25% medium in the reaction (**Fig. 2a**). However, clinical samples contain additional material from the swab and other unknown and potentially inhibitory agents. In addition, due to potentially large variability between clinical samples, it is important to characterize inhibition curves in multiple individual clinical samples rather than an averaged mix of samples. The optimal amount of sample input in hid-RT-qPCR should be a balance between possible inhibition from the sample and the amount of RNA going into the reaction. To identify the optimal range of sample input in clinical samples, we performed dilution series (10 to 0.01μl) of individual COVID-19 positive heat-inactivated nasopharyngeal swab samples, and found an input of 1–4μl sample in a 20μl RT-qPCR reaction to be optimal, avoiding the sharp inhibitory effect at higher amounts of sample input observed in some individual samples, yet minimizing CT (**Fig. 2f-g**). To begin to explore whether direct RT-qPCR on heat-inactivated samples might allow effective COVID-19 diagnostics, we performed heat inactivation (65°C 30min) of frozen (−20°C) aliquots from 85 clinically diagnosed nasopharyngeal samples and performed hid-RT-qPCR blindly to their COVID-19 status. We used 4μl input and primers N1, RdRP, as well as RNase P for assessment of sample integrity. Thereafter, we combined the results of hid-RT-qPCR with CT values from the clinical diagnostics performed on extracted RNA (MagNA Pure 96, Roche Diagnostics, test targeting E and RdRP). We observed strong correlation between CT values of extracted and heat-inactivated samples (**Fig. 2h**) and overall agreement of SARS-CoV-2 calls (**Fig. 2i**). However, CT values for hid-RT-qPCR (65°C 30min) on frozen samples were higher than for fresh RNA eluates of the same samples (median 6.7 CT difference) (**Fig. 2h-i**). This was expected given that *(1)* more RNA was loaded for eluates (2.5*x*, standard 10μl input *vs*. 4μl volume equivalent in hid-RT-qPCR), *(2)* RNA extraction of eluates was performed on fresh samples while the aliquots used for hid-RT-qPCR had been frozen and stored at –20°C before heat inactivation, *(3)* heating may degrade RNA in presence of RNases and/or metal ions (metal-ion-based RNA cleavage). By performing RNA re-extraction from 19 freeze-thawed aliquots and comparing CT values to eluates of matched fresh aliquots of the same nasopharyngeal specimens we found the effect of freeze-thaw to result in +2–3 CT (**Fig. 2j**). In summary, our data from 85 clinically diagnosed nasopharyngeal frozen samples showed that hid-RT-qPCR could be a working option to extraction-based of SARS-CoV-2 diagnostics, and that efforts to optimize the hid-RT-qPCR protocol for maximum performance would be worthwhile. ### Optimisation of hid-RT-qPCR To identify an optimal heat-inactivation program preceding hid-RT-qPCR, we subjected 50μl fresh (non-frozen) aliquots of the same clinical nasopharyngeal samples in transport medium to different temperatures and incubation times (65°C 30min; 95°C 5min; 95°C 10min; 95°C 15min; and 98°C 5min; n≥ 11). We observed consistent improvement (reduction) of hid-RT-qPCR CT values in aliquots inactivated at 95°C, 5 min compared to 65°C, 30 min (median CT change: –1.3, *P*= 1.1×10−5, Wilcoxon signed-rank test, N1 primer-probe set) (**Fig. 3a**). Interestingly, all the other high-temperature (≥95°C) conditions tested resulted in similar CT as the 95°C 5min treatment (*P*> 0.05, FDR-corrected Wilcoxon signed-rank tests) (**Fig. 3a**). Thus, we conclude that inactivation before hid-RT-qPCR should be performed at 95–98°C. This result is fortunate and important, since incubation at such high temperature should completely inactivate the virus [11]. Moreover, the stability across incubation times (5–15 min) at high temperature demonstrates a remarkable robustness of the procedure. ![Figure 3.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/12/2020.04.17.20067348/F3.medium.gif) [Figure 3.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/F3) Figure 3. Optimisation of heat-inactivation conditions for SARS-CoV-2 hid-RT-qPCR (**a**) CT values for aliquots of nasopharyngeal swab samples inactivated at different temperature and time conditions, shown on absolute scale (left) and CT change relative to the 95°C 5 min condition (right). *P*-values calculated as two-tailed Wilcoxon signed rank tests. *FDR corrected for multiple testing. Median values shown in red. (**b**) Heatmaps of ranks based on absolute CT values for nasopharyngeal swab samples (n = 20) with or without addition of polyvinylsulfonic acid (PVSA) and/or ethylenediaminetetraacetic acid (EDTA) before performing heat inactivation at 95°C for 5 minutes. For EDTA-containing samples, equimolar supplement of MgCl2 was added to the RT-qPCR reaction as indicated. The same samples were evaluated by RT-qPCR repeatedly at different number of days stored at 4°C after the heat activation. All conditions are listed in Table 2. (**c**) Boxplots of CT values relative to the 95°C condition without additives, ordered by the rank in (b). (**d**) Change in rank over days of storage at 4°C. Colors and order same as in (b-c). Annotations refer to samples stored for 0 days. (**e**) Change in CT values over days stored at 4°C across different conditions. Thick line and ribbon indicate fitted LOESS curve and ±95% confidence interval, respectively. The dashed line indicates the 95°C (without additive) condition. Next, we tested whether hid-RT-qPCR could be improved by addition of the chemical, thermostable, RNase inhibitor polyvinylsulfonic acid (PVSA), and/or the chelating agent ethylenediaminetetraacetic acid (EDTA) during heat inactivation. PVSA halves *in vitro* RNase A activity (50% inhibition, IC50) at the concentration 150μg/ml and halves the RNase activity of E. coli lysate (IC50) at 430μg/ml [12]. At the same time, PVSA might inhibit RT-qPCR. By performing dilution series of PVSA using synthetic SARS-CoV-2 RNA (SKU102024-MN908947.3, Twist Biosciences) as template, we identified a concentration range with prospective RNase inhibition in the sample yet limited inhibition of RT-qPCR (**Supplementary fig. 1d**). We then supplemented fresh aliquots of 20 COVID-19-diagnosed clinical nasopharyngeal samples in transport medium with various amounts of PVSA and/or EDTA, and for EDTA-containing conditions we further performed tests supplementing equimolar amounts of MgCl2 in the RT-qPCR mix (12 conditions and 20 samples, n= 240) (**Table 2**). To additionally test the RNA stability in different treatments over time in storage (4°C) we determined the SARS-CoV-2 RT-qPCR CT change in the same samples over different number of days (0, 1, 4 and 7) after heat inactivation. We ranked the treatments within each sample and day, and we observed that 95°C 5min + 150μg/ml PVSA produced the highest score (lowest CT) followed by 95°C 5min without additives across day 0 to 4 (**Fig. 3b-d** and **Table 2**). The benefit of EDTA and higher concentrations of PVSA only became apparent after 4–7 days in storage (4°C) (**Fig. 3d-e** and **Table 2**). Interestingly, the results of the time series showed that heat-inactivated samples are surprisingly stable up to one week in storage at 4°C. Given the good performance of the 95°C 5min condition without additives (less than 1 CT difference compared 150μg/ml PVSA, **Fig. 3c**) together with its simplicity in sample handling, we selected this condition for further benchmarking. View this table: [Table 3.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/T3) Table 3. Accuracy, sensitivity and specificity of SARS-CoV-2 hid-RT-qPCR (95°C 5 min) using the Roche cobas 6800 analyzer as diagnostic reference. Each parameter was calculated using the cobas diagnostic call as reference, either requiring both cobas primer-probe sets producing signal (CT≤40) to call a sample COVID-19 positive (i.e. in accordance to themanufacturer’s specification), or, requiring only one (any) cobas primer-probe set producing signal to call a sample positive. As comparison, the parameters were calculated also for the cobas ORF1 primer-probe set, using the more sensitive cobas E primer-probe set as the reference (parameters calculated using the same samples as for hid-RT-qPCR). Primer-probesets used in the comparisons within brackets. Binominal test *P*-values. N= 597 fresh nasopharyngeal swab samples. ### Validation of hid-RT-qPCR against standardised clinical testing Having optimised the heat-inactivation conditions, we next benchmarked SARS-CoV-2 hid-RT-qPCR (95°C 5 min) using the cobas 6800 system (Roche Diagnostics; hereby referred to as cobas) and a large set of paired clinical nasopharyngeal swab samples as reference. To first test the cobas performance compared to conventional RT-qPCR on eluated RNA, we determined CT values of 21 purified clinical nasopharyngeal swab samples and performed a limit-of-detection experiment with the same sample (dilution from 1:100 to 1:100,000) on both systems, and observed a higher rate of detection and sensitivity for cobas (**Supplementary table 1–2**). Given its performance and the fact that the cobas is a standardized and fully automated system (avoiding manual sample handling) we deemed it to be a suitable system for validation of SARS-CoV-2 hid-RT-qPCR. We collected aliquots of 597 clinical nasopharyngeal swab samples diagnosed on the cobas analyser the same day using two targets (primer-probe sets towards ORF1 and E; “Target 1” and “Target 2” cobas SARS-CoV-2, P/N: 09175431190, Roche Diagnostics) and performed heat inactivation (95°C, 5 min) followed by SARS-CoV-2 hid-RT-qPCR (N1 primer-probe set, 4μl sample input). CT values of hid-RT-qPCR and cobas correlated well and had similar CT value distributions (*P*=0.11 and 0.88; N1 *vs*. ORF1 and E, respectively; Kolmogorov–Smirnov test) (**Fig. 4a-b**). In addition to the 597 samples cross-compared with hid-RT-qPCR, we plotted 9437 historical nasopharyngeal SARS-CoV-2 CT values collected on the same cobas machine and observed that the CT value distribution of the 597 samples was representative of the larger set (*P*=0.23 and 0.35; ORF1 and E, respectively; Kolmogorov–Smirnov test) (**Fig. 4c-d**). Finally, we classified the COVID-19 status of the 597 nasopharyngeal swab samples on the cobas, either requiring positive signal (CT≤ 40) for both targets (ORF1 and E), or, any target (ORF1 and/or E) to call a sample SARS-CoV-2 positive (**Fig. 4e**, leftmost bars). We plotted a heatmap of SARS-CoV-2 detection (CT) and observed remarkable agreement between cobas and hid-RT-qPCR (**Fig. 4e-g**). Using the diagnostic call of both cobas targets as reference, hid-RT-qPCR had an accuracy, sensitivity and specificity of 98.8% (95% confidence interval, CI95: 97.5–99.5%), 96.0% and 99.8%, respectively (**Table 3**). Requiring only one cobas target to call samples SARS-CoV-2 positive, the hid-RT-qPCR accuracy, sensitivity and specificity were 97.5% (CI95: 95.9–98.6%), 91.4% and 99.8%, respectively. We observed the performance of hid-RT-qPCR (95°C 5 min) using the N1 primer-probe set to be similar to the cobas ORF1 target (**Table 3** and **Fig. 4f-h**). Next, we calculated the sensitivity of hid-RT-qPCR at different CT thresholds and CT bins (as detected by the most sensitive cobas E-gene target; limit of detection ORF1: 0.009 TCID50/ml, E-gen: 0.003 TCID50/ml; **Supplementary table 2**) and found that hid-RT-qPCR effectively only lost in sensitivity for samples detected at CT 35–40 by the cobas E target, i.e. close to the limit of detection (**Fig 4i-j**). View this table: [Supplementary table 1.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/T4) Supplementary table 1. CT values for 21 clinical nasopharyngeal swab samples quantified using cobas 6800 and TaqPath SARS-CoV-2 RT-qPCR on eluted RNA. We observed dropout of detection in four positive samples in RT-qPCR on eluted RNA.ND: Not detected (CT >40) ViCu: Virocult. TSw: Transwab. View this table: [Supplementary table 2.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/T5) Supplementary table 2. Limit-of-detection data. CT values for a COVID-19 positive clinical nasopharyngeal swab sample diluted1:100, 1:1000, 1:10,000 or 1:100,000 in PBS and quantified using cobas 6800 and TaqPath SARS-CoV-2 RT-qPCR on eluted RNA. ![Figure 4.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/12/2020.04.17.20067348/F4.medium.gif) [Figure 4.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/F4) Figure 4. Validation of SARS-CoV-2 hid-RT-qPCR (**a**) Scatter plots of CT values from clinical diagnostics performed on the Roche Diagnostics cobas 6800 (y-axis) and hid-RTq-PCR (x-axis) for 597 nasopharyngeal swab samples, shown for different primer-probe set comparisons. Rho indicates Spearman correlation of positive samples. ND: not detected. (**b**) Histograms of CT values from 597 paired nasopharyngeal swab samples shown for the following primer-probe sets and conditions: cobas ORF1, cobas E, and hid-RT-qPCR N1. ND: not detected. (**c**) Histograms of CT values from 9437 clinical samples analysed on a cobas 6800, show for ORF1 (left) and E (right). (**d**) Scatter plots of CT values from 9437 clinical samples analysed on a cobas 6800. Rho indicates Spearman correlation of positive samples. ND: not detected. (**e**) Heatmap of CT values from diagnostics performed on 597 clinical nasopharyngeal swab samples using the cobas 6800 (ORF1 and E primer-probe sets) and hid-RT-qPCR (N1). The bars to the left indicate the clinical call from the cobas diagnostics, considering detection either in both or only one (any) primer-probe set for the diagnostic call. (**f-h**) Confusion matrix of diagnostic call from cobas 6000 and hid-RT-qPCR (N1) for the data shown in (e). (**i**) Percent sensitivity of hid-RT-qPCR (N1) using cobas ORF1 and E primer-probe sets as reference, shown as a function of CT threshold for the same sample as detected by the cobas primer-probe set E. The red lines denote the cumulative sensitivity below the given CT threshold and the bars denote the sensitivity in 5-CT bins. (**j**) Same as (i) but using cobas ORF1 and/or E primer-probe sets as reference for calculating sensitivity. Together, these data demonstrate that a simple heat inactivation program followed by direct RT-qPCR using the CDC primer-probe pair N1 detects SARS-CoV-2 with remarkable accuracy, sensitivity and specificity given the ease of the method. As a result, RNA-purification-free SARSCoV-2 detection is viable, enabling cheap, scalable and rapid testing of COVID-19. ### Direct RT-qPCR on lysed SARS-CoV-2 samples SARS coronavirus envelopes are self-assembled particles in which the lipid bilayer is a weak spot [13], thus the viral envelope can be ruptured by surfactants and at the same time viral RNA can be released from similarly lysed human cells in the sample [14]. A direct route for SARS-CoV-2 screening could be self-testing using nose and throat swabs, or even on saliva, followed by lysis directly before RT-qPCR on unpurified samples. RT-qPCR assays directly on detergent-inactivated samples would require an RT-qPCR assay resilient to high concentrations of detergent. We monitored the effect of Triton X-100 and Tween-20 on SARS-CoV-2 RT-qPCR using spike-in of 50,000 copies of synthetic full-genome SARS-CoV-2 RNA (SKU102024-MN908947.3, Twist Biosciences) and the N1 primer-probe set. CT values were only modestly affected (+1–2 CT) when incubated with as much as 5% Triton X-100 or 10% Tween-20 in the RT-qPCR reaction (**Fig. 5a-b**). We observed lowered levels of fluorescence in the plateau phase in qPCR at increased concentrations of detergent, without markedly affecting the CT (**Fig. 5c-d**). To test whether actual SARS-CoV-2 RNA could be detected after direct lysis, we obtained six aliquots of saliva and six combined nose and throat swabs in PBS from six deidentified donors (Methods), of which four had been identified as COVID-19 positive and two as negative in extraction-based routine diagnostics (sample aliquots obtained from the Public Health Agency of Sweden). Notably, these samples were not collected by health care professionals using clinical grade flocked plastic swabs, rather the samples were self-collected using simple cotton swabs (deposition in PBS) and a jar without storage buffer for saliva (Methods). Furthermore, at the time of our experiment, these samples had been frozen, thawed and stored at room temperature for several hours combined. We tested these samples blindly, by mixing 5μl sample (saliva or nose + throat swab) with 5μl of 10 or 20% Triton X-100 and performed the SARS-CoV-2 RT-qPCR using the N1 primer-probe set directly on these lysates. Indeed, all four COVID-19 positive donor individuals were correctly called as SARS-CoV-2 positive in at least one Triton X-100 condition or sample (saliva and/or nose + throat swab), while negative controls lacked signal (**Fig. 5e**). ![Figure 5.](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2020/05/12/2020.04.17.20067348/F5.medium.gif) [Figure 5.](http://medrxiv.org/content/early/2020/05/12/2020.04.17.20067348/F5) Figure 5. Direct SARS-CoV-2 RT-qPCR detection from lysate (**a-b**) Line charts of CT values (y-axis) from RT-qPCR performed with different percent (vol./vol.) Triton X-100 (a) or Tween-20 (b) (x-axis) in the reaction. (**c-d**) Amplification plots showing normalized reporter value (ΔRn, linear scale) as a function of qPCR cycle for the experiments and samples described in (a-b). (**e**) Barplots of CT from SARS-CoV-2 RT-qPCR using the N1 primer-probe set performed directly on lysed donor samples after storage and freeze thaw from self-sampling (saliva or nasal + throat swab suspensions taken with cotton tipped wooden sticks) without purification. Percent Triton X-100 indicates the percentage detergent in the sample (half concentration in the RT-qPCR reaction). Crosses indicate CT values from diagnostics performed on extracted RNA from fresh aliquots. These initial but principal results demonstrate that direct SARS-CoV-2 RT-qPCR can also be applied on detergent-inactivated self-sampled material, opening an alternative route for large-scale population screening. ## DISCUSSION Scalable, rapid and affordable COVID-19 diagnostics could help to limit the spread of SARS-CoV-2, consequently saving lives. However, RNA extraction constitutes a barrier to scale-up of testing. We explored procedures to circumvent RNA extraction by performing RT-qPCR directly on heat-inactivated subject samples and sample lysates. Our results show that RT-qPCR-based testing for SARS-CoV-2 infection can be performed through significantly simpler protocols and without the use of RNA extraction kits, nor other special kits. The results also show that this can be achieved without major sacrifice in accuracy of determining negative and positive cases. The procedure could be especially useful for massively scaling up SARS-CoV-2 testing, as the logistics and cost of RNA purification could be unworkable in mass testing. Importantly, the direct method is also attractive in settings where repeated, cheaper, and quicker testing is desirable, for example in frequent testing of healthcare personnel. The direct method that we present would also be compatible with downstream sequencing-based detection as an alternative to qPCR. We determined RT-qPCR inhibition profiles of different transport media as well as the optimal amount of reaction input of nasopharyngeal swab samples (**Fig. 2**). Further effort should be invested in similar characterization of many more brands and types of transport media in circulation. We propose that characterization of RT and PCR inhibition should become a standard (requirement) for commercial transport media in the future, as to assist direct testing in forthcoming epidemics. In the ongoing COVID-19 pandemic, we question whether it is at all suitable to deposit COVID-19 samples in conventional transport medium. Instead, we suggest that the swab material could be collected in a generic buffer that does not inhibit RT-qPCR, especially as downstream viral culturing in lab is not meaningful for the vast majority of samples. Heat inactivation cleaves RNA into shorter fragments, and thus primer and probe considerations in hid-RT-qPCR should be more important for its sensitivity than for extraction-based RT-qPCR of more intact RNA. Accordingly, we observed that the primer-probe set with the shortest amplicon (N1, 72bp) performed best in hid-RT-qPCR and the longest amplicon (E, 113bp) performed the worst (**Fig. 2d-e**). In conclusion, short amplicon targets should be used in hid-RT-qPCR. After optimizing heat-inactivation conditions (**Fig. 3**), we validated hid-RT-qPCR using the standardized and sensitive cobas 6800 system (Roche Diagnostics) as reference method (**Fig. 4**) and we observed high accuracy, sensitivity and specificity of hid-RT-qPCR (**Table 3**). We propose that the sensitivity of hid-RT-qPCR could be improved even more if swab samples were to be collected in smaller volumes of a buffer that do not inhibit RT-qPCR (such as TE) as the input amount of RNA could then be increased. Following this, we hypothesized that the direct RT-qPCR pipeline for COVID-19 testing might also be implemented by sampling directly into a lysis buffer containing detergents such as Triton X-100. Lysed samples could immediately be subjected to RT-qPCR analysis and diluted in the RT-qPCR master mix without intermediate steps. Although the data is limited and the procedure should be refined, our results on self-collected saliva and nose/throat samples using cotton swabs (**Fig. 5e**) show that this strategy is workable. If the sensitivity ultimately proves to be adequate for meaningful decisions on self-isolation to limit spread, then this method could be applied to samples taken by the test-subjects themselves, allowing massive screening of the population. ## METHODS *Sample collection:* Clinical samples (nasopharyngeal swabs) were collected and deposited in transport medium (Virocult, Transwab or Eswab) at the Karolinska University Hospital, Stockholm, Sweden. For routine diagnostics, 200µl of sample in transport medium was inactivated by addition of 200µl MagNA Pure 96 External Lysis Buffer (06374913001, Roche Diagnostics), and then extracted using MagNA Pure 96 DNA and Viral NA SV Kit (06543588001, Roche Diagnostics) or MGI Easy Magnetic Beads Virus DNA/RNA Extraction Kit (1000006989, MGI). The samples used in direct lysis experiment (**Fig. 5e**) were self-collected volunteer sample from a COVID-19 clinical screen performed in the Stockholm area, organized by the Public Health Agency of Sweden. For swab samples (nose + throat), two cotton tipped wooden swabs were used. Written instructs were provided to introduce one swabs into the throat via the chin as far back in the throat as possible and scrape for 10-20s then rinse the swab in the provided buffer for 10-15s. Instructions were also given to take a second wooden swab and introduce into the nose and scrape for 10-20s in each nostril, followed by a 10-15s rinse in the same buffer tube as the throat swab. The buffer in the swab test was 100mM PBS pH 7.4. Further instructions were to leave a saliva sample at the same time by spitting 3-4 times in a small jar during a 5-10min period. The samples were picked up and transported to a laboratory for testing typically within 1-10hrs after the sampling. Samples were stored at +4°C and RNA was extracted within 24h and tested using RT-qPCR. For the current study, samples were deidentified aliquots of these same samples that had been subsequently frozen at −20°C for approximately 7 days, thawed and kept at room temperature ~2hrs, before performing the direct lysis experiments described herein. In this work we also used anonymized or pseudo-anonymized surplus material from samples that had been collected for clinical diagnostics of SARS-CoV-2. Informed consent was not obtained from the patients. This is in accordance with the Swedish Act concerning the Ethical Review of Research Involving Humans, which allows development and improvement of diagnostic assays using patient samples which were collected to perform the testing in question. *Heat inactivation:* Nasopharyngeal swab samples in approximately 1ml transfer medium were vortexed and 50µl aliquots of each sample were transferred to 96-well PCR plates which were sealed (adhesive aluminum foil, VWR cat. 60941–112) and subjected to thermal inactivation using a thermal cycler with heated lid and using a thermal sealing mat. Alternatively, approximately 200µl aliquots were inactivated in 1.5ml Eppendorf tubes using a heat block. Before RT-qPCR testing, the plates were centrifuged to collect debris in the bottom of the wells, and 4 µl sample for RT-qPCR were collected from the liquid upper phase using a 10 µl multi pipette and added to plates containing 16µl TaqPath mastermix. Each time the seal of a plate was opened we replaced the seal with a new seal to avoid cross contamination. *One-step RT-qPCR:* For reverse transcription and quantitative PCR we used the one-step TaqPath RT-qPCR master mix (Thermo, A15299) according to manufacturer’s instructions. Final reactions of 20µl were formed by mixing 5µl TaqPath master mix, primers-probe mix, sample and RNase free water to fill the reaction. Primer and probe concentrations in the RT-qPCR reaction were as follows, E and RdRP: 300nM each primer, 200nM probe; N1 and RdRP: 500nM each primer, 125nM probe. Primer and probe sequences are listed in **Table 1**. The thermal cycling steps were: 25°C for 2 min, 50°C for 15min, 95°C for 2min, and 45 cycles of 95°C for 3s and 56°C for 30s. RT-qPCR was performed on a Step-One-Plus real time PCR machine (Applied Biosystems) using the StepOne Software v2.3. The samples from the self-test screen were subjected to the same protocol as described above but without heat inactivation. Briefly, samples (swab samples in PBS or pure saliva) were mixed with equal volume of 10% or 20% Triton X-100 at room temperature (approximately 5min) before dilution 1:1 in a 10µl TaqPath RT-qPCR master mix as described above. Clinical COVID-19 diagnostics at the Karolinska University Hospital, Stockholm, Sweden were similarly performed using TaqPath and primer-probe sets for E and RdRP and 10µl RNA eluate as sample. *Two-step RT-qPCR:* Reverse transcription was performed by mixing subject sample (4µl in case of clinical nasopharyngeal swab sample), 1µl 10mM dNTPs, 0.15µl 50µM random hexamers (N8080127, Thermo), 0.1µl RNase inhibitor (2313B, TaKaRa), 0.4µl 0.5% Triton X-100 and RNase free water up to 4.5µl, followed by incubation at 72°C for 3min. The sample was placed on ice and a mix containing 0.5µl 100mM DTT, 2µl 5M betaine, 0.1µl 1M MgCl2, 0.25µl RNase inhibitor (2313B, TaKaRa), 2µl 5x Superscript II buffer, 0.5µl Superscript II (Invitrogen) and water up to 5.5µl was added. The samples were then incubated at 25°C for 10min followed by 42 °C for 25min and finally for 70°C for 15min. Amplification of 10µl cDNA (RT mix) using primers and probes described in **Table 1** was performed using BioTaq DNA polymerase (Bio-21040, Bioline) in a 20µl reaction containing 2µl 10x NH4 Reaction Buffer (Bioline), 1.2µl 50mM MgCl2, 0.2µl 100mM dNTP Mix and water up to 20 µl. The thermal cycling steps were: 25°C for 2min, 50 °C for 15min, 95°C for 2min, and 45 cycles of 95°C for 3s and 56°C for 30s. qPCR was performed on a Step-One-Plus real time PCR machine (Applied Biosystems) using the StepOne Software v2.3. *Cobas 6800:* The cobas 6800 is a fully automated instrument that once samples have been loaded performs extraction, amplification and detection. For details regarding the assay see Roche Diagnostics document 09179909001–01EN (Doc Rev. 1.0). Roche supplied reagents specific for the cobas SARS-CoV2 analysis are cobas® SARSCoV-2 (P/N: 09175431190) cobas® SARS-CoV-2 Control Kit (P/N: 09175440190) cobas® 6800/8800 Buffer Negative Control Kit (P/N: 07002238190). Prior to analysis MagNA Pure 96 External Lysis Buffer (06374913001, Roche Diagnostics) were added to all samples at a ratio of 1:1 resulting in 150 ul of each patient samples being analysed. The instrument software version used was 01.03.08.1011. *Gel electrophoresis*: Products of RT-qPCR were separated by electrophoresis on a 3% agarose gel in 1xTBE buffer. The lengths of the products were determined relative to an Ultra-low Range DNA ladder (Thermo, SM1213). Images were taken using an Imagequant Las4000 camera system (Cytiva). *Data analysis*: Graphical representations of data and statistical testing were performed using R. Effect of optimization parameters over time was modelled using linear regression using CT relative to day 0 as the response variable and time (days), EDTA, PVSA and Mg2+ additions as predictors both alone and as interaction terms with time. Accuracy, sensitivity and specificity was calculated using the confusionMatrix function in the caret R package [15]. ## Data Availability All data are available. ## Ethics statement The intent of the work was clinical methods development, as a response to the COVID-19 emergency. In this work we used anonymized or pseudo-anonymized surplus material from samples that had been collected for clinical diagnostics of SARSCoV-2. This is in accordance with the Swedish Act concerning the Ethical Review of Research Involving Humans, which allows development and improvement of diagnostic assays using patient samples which were collected to perform the testing in question. ## Author Contributions IS, ME, NP, JoAa, HS and BR performed experiments. AL and BR analyzed data and prepared figures. ME, MV, SM, JaAl, BH and BR participated in sample handling and organization. BR supervised the study and wrote the manuscript. All authors participated in manuscript editing and approved the manuscript. ## Competing Interests The authors declare no competing interests. ## Acknowledgements This research was supported and funded by the SciLifeLab COVID-19 initiative, the Knut and Alice Wallenberg Foundation (KAW 2020.0182), the Swedish Research Council (2017–01723) and the Ragnar Söderberg Foundation to BR. We are grateful to Sten Linnarsson for providing synthetic full-genome SARS-CoV-2 RNA and to members of the Reinius lab for comments and discussions. * Received April 17, 2020. * Revision received May 7, 2020. * Accepted May 12, 2020. * © 2020, Posted by Cold Spring Harbor Laboratory The copyright holder for this pre-print is the author. All rights reserved. The material may not be redistributed, re-used or adapted without the author's permission. ## REFERENCES 1. 1.Taipale J, Romer P, Linnarsson S. Population-scale testing can suppress the spread of COVID-19. medRxiv. May 2020:2020.04.27.20078329. (2020) doi:10.1101/2020.04.27.20078329 [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoibWVkcnhpdiI7czo1OiJyZXNpZCI7czoyMToiMjAyMC4wNC4yNy4yMDA3ODMyOXYyIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjAvMDUvMTIvMjAyMC4wNC4xNy4yMDA2NzM0OC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 2. 2.Ferguson N. M et al. Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand. (2020) [https://spiral.imperial.ac.uk:8443/bitstream/10044/1/77482/8/2020-03-16-COVID19-Report-9.pdf](https://spiral.imperial.ac.uk:8443/bitstream/10044/1/77482/8/2020-03-16-COVID19-Report-9.pdf) 3. 3.Corman, V. M. et al. Assays for laboratory confirmation of novel human coronavirus (hCoV-EMC) infections. Eurosurveillance 17, 20334 (2012) 4. 4.Corman, V. M. et al. Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR. Euro Surveill. 25, (2020). 5. 5.Centers for Disease Control and Prevention. Real-Time RT-PCR Panel for Detection 2019-Novel Coronavirus. (2020) [https://www.cdc.gov/coronavirus/2019-ncov/downloads/rtpcr-panel-for-detection-instructions.pdf](https://www.cdc.gov/coronavirus/2019-ncov/downloads/rtpcr-panel-for-detection-instructions.pdf) 6. 6.Centers for Disease Control and Prevention. Information for Laboratories: 2019-nCoV | CDC. Acceptable Commercial Primers and Probes (2020). [https://www.cdc.gov/coronavirus/2019-ncov/lab/index.html](https://www.cdc.gov/coronavirus/2019-ncov/lab/index.html). 7. 7.Merindol N. et al. Optimization of SARS-CoV-2 detection by RTQPCR without RNA extraction. (2020) [https://www.biorxiv.org/content/10.1101/2020.04.06.028902v1.full.pdf](https://www.biorxiv.org/content/10.1101/2020.04.06.028902v1.full.pdf) 8. 8.Bruce, E. A. et al. Rt-Qpcr Detection Of Sars-Cov-2 Rna From Patient Nasopharyngeal Swab Using Qiagen Rneasy Kits Or Directly Via Omission Of An Rna Extraction Step. bioRxiv 2020.03.20.001008 (2020) doi:10.1101/2020.03.20.001008 [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoiYmlvcnhpdiI7czo1OiJyZXNpZCI7czoxOToiMjAyMC4wMy4yMC4wMDEwMDh2MiI7czo0OiJhdG9tIjtzOjUwOiIvbWVkcnhpdi9lYXJseS8yMDIwLzA1LzEyLzIwMjAuMDQuMTcuMjAwNjczNDguYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 9. 9.Fomsgaard, A. S. and Rosenstierne, M. W. An alternative workflow for molecular detection of SARS-CoV-2 – escape from the NA extraction kit-shortage, Copenhagen, Denmark, March 2020. Eurosurveillance 25, 2000398 (2020) 10. 10.Barra, G. B. et al. Analytical sensibility and specificity of two RT-qPCR protocols for SARS-CoV-2 detection performed in an automated workflow. medRxiv 2020.03.07.20032326 (2020) 11. 11.Chin A. W. H. et al. Stability of SARS-CoV-2 in different environmental conditions. The Lancet Microbe. 2020;(). doi:10.1016/S2666-5247(20)30003-3 [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/S2666-5247(20)30003-3&link_type=DOI) 12. 12.Earl C. C. et al. Polyvinylsulfonic acid: A Low-cost RNase inhibitor for enhanced RNA preservation and cell-free protein translation. Bioengineered. 2018;9(1):90–97. doi:10.1080/21655979.2017.1313648 [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1080/21655979.2017.1313648&link_type=DOI) 13. 13.Khattari, Z. et al. SARS Coronavirus E Protein in Phospholipid Bilayers: An X-Ray Study. Biophys. J. 90, 2038–2050 (2006). [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1529/biophysj.105.072892&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=16361349&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F12%2F2020.04.17.20067348.atom) 14. 14.Rabenau, H. F., Kampf, G., Cinatl, J. & Doerr, H. W. Efficacy of various disinfectants against SARS coronavirus. J. Hosp. Infect. 61, 107–111 (2005) [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jhin.2004.12.023&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=15923059&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F12%2F2020.04.17.20067348.atom) 15. 15.Kuhn M. Building Predictive Models in R Using the caret Package. J Stat Softw. 2008;28(5):1–26. doi:10.18637/jss.v028.i05 [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.18637/jss.v028.i05&link_type=DOI) [PubMed](http://medrxiv.org/lookup/external-ref?access_num=27774042&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2020%2F05%2F12%2F2020.04.17.20067348.atom)