key: cord-1036085-nk2iae6b authors: Zheng, Yang; Jin, Jun; Shao, Ziqiang; Liu, Jingquan; Zhang, Run; Sun, Renhua; Hu, Bangchuan title: Development and clinical validation of a droplet digital PCR assay for detecting Acinetobacter baumannii and Klebsiella pneumoniae in patients with suspected bloodstream infections date: 2021-11-16 journal: Microbiologyopen DOI: 10.1002/mbo3.1247 sha: 3683e274c289f887483a0e26d20a11156db68075 doc_id: 1036085 cord_uid: nk2iae6b The relatively long turnaround time and low sensitivity of traditional blood culture‐based diagnosis may delay effective antibiotic therapy for patients with bloodstream infections (BSIs). A rapid and sensitive pathogen detection method is urgently required to reduce the morbidity and mortality associated with BSIs. Acinetobacter baumannii and Klebsiella pneumoniae are two major microorganisms that cause BSIs. Here we report a novel droplet digital polymerase chain reaction (ddPCR) assay that can detect A. baumannii and K. pneumoniae in blood samples within 4 h, with a specificity of 100% for each strain and a limit of detection at 0.93 copies/μl for A. baumannii and 0.27 copies/μl for K. pneumoniae. Clinical validation of 170 patients with suspected BSIs showed that compared to blood cultures that detected four (2.4%) A. baumannii cases and seven (4.1%) K. pneumoniae cases, ddPCR detected 23 (13.5%) A. baumannii cases, 26 (15.3%) K. pneumoniae cases, and four (2.4%) co‐infection cases, including the 11 cases detected via blood culture. In addition, patients who tested positive via ddPCR alone (n = 42) had significantly lower serum concentrations of procalcitonin and lactate, SOFA and APACHE II scores, and 28‐day mortality than those reported positive via both blood culture and ddPCR (n = 11), suggesting that patients with less severe symptoms can potentially benefit from ddPCR‐based diagnosis. In conclusion, our study suggests that ddPCR represents a sensitive and rapid method for identifying causal pathogens in blood samples and guiding treatment decisions in the early stages of BSIs. Bloodstream infections (BSIs) represent a major cause of death worldwide, contributing to increases in healthcare costs, length of hospital stay, and in-hospital morbidity (McNamara et al., 2018) . Timely and accurate pathogen identification is critical to select the appropriate antimicrobial treatment for patients in the early stages of BSI. Blood culture remains the gold standard for identifying pathogens in BSIs (Blevins & Bronze, 2010) . However, it is limited by its low sensitivity and long turnaround time (Riedel, & Carroll, 2016; Tabak et al., 2018) . In patients with sepsis within the first 6 h of documented hypotension, every 1-h delay in an appropriate antibiotic therapy leads to an average increase in the mortality rate of 7.6% (Kumar et al., 2006) . For hospitalized patients with bacterial infections, inappropriate initial antimicrobial treatment almost doubles the risk of 30-day mortality (Fraser et al., 2006) . Thus, it is necessary to develop a rapid and accurate method for identifying causal pathogens in BSIs. Culture-independent, real-time polymerase chain reaction (PCR)based, or microarray-based methods, such as SeptiFast (Roche), Magicplex (Seegene), and TaqMan array card assay (Academy of Military Medical Science, Beijing, China), show promise in rapidly identifying pathogens and guiding early targeted antibiotic therapy in BSIs. However, their low sensitivities, ranging from 29% to 79.4%, may limit their clinical application (Buehler et al., 2016; Riedel, & Carroll, 2016; Warhurst et al., 2015; Zboromyrska et al., 2019; Zhang et al., 2018) . Recently, droplet digital PCR (ddPCR) has been developed as a novel molecular diagnostic technique for the accurate detection and absolute quantitation of nucleic acids without the need of generating a calibration curve . In ddPCR, the template is separated into thousands of nanolitre-sized droplets and amplified. After amplification, the number of positive and negative reactions is counted, and the copy number of the template is calculated using Poisson statistics. So, ddPCR has been increasingly applied as a versatile tool with high sensitivity, accuracy, and precision in multiple clinical scenarios, including oncology (Postel et al., 2018) , non-invasive prenatal testing (Tan et al., 2019) , and diagnosis of infectious diseases (Kelley et al., 2013; Pholwat et al., 2013; Sedlak, Cook, Huang, et al., 2014) . Furthermore, few studies report that ddPCR is superior to conventional methods in its application for the detection of infectious pathogens, including bacteria, fungi, and viruses Park et al., 2021; Wouters et al., 2020) . Circulating cell-free (cf) DNA molecules originate from dying cells and from colonizing or invading microorganisms, which release DNA fragments into the blood as they break down, and their detection has emerged as a powerful non-invasive diagnostic tool in the fields of prenatal screening, transplantation, and oncology (Aravanis et al., 2017; Bloom et al., 2017; Fan et al., 2008 ). An increasing number of studies have also indicated that microbial cfDNA detection may enable the reliable identification of various infections, such as invasive fungal infection, tuberculosis, and sepsis (Han et al., 2020) . Grumaz et al. observed significantly increased levels of cfDNA derived from pathogenic bacteria in the plasma of patients with sepsis, and subsequently reported that microbial cfDNA-based next-generation sequencing (NGS) can provide higher sensitivity and specificity than that obtained via blood culture for the diagnosis of sepsis (Grumaz et al., 2016 (Grumaz et al., ,2019 . Moreover, in a case with confirmed Acinetobacter baumannii BSI, Liao et al. found that 1 ml of plasma contains more than 1000 copies of the A. baumannii genome, which exceeds the microbial load of intact live bacteria in the majority of blood samples derived from patients with sepsis (Liao et al., 2020) . Thus, detection of microbial cfDNA may provide higher sensitivity for the diagnosis of BSIs than that of detecting intact live pathogens in blood samples. Additionally, in comparison with microbial cfDNA, the direct detection of pathogenic DNA in whole blood may be limited by the high quantity of human DNA that interferes with primer and probe binding in species-specific PCR assays. A. baumannii and Klebsiella pneumoniae represent two major Gram-negative bacteria involved in BSIs and show a high capability to develop antibiotic resistance. BSIs caused by multidrug-resistant A. baumannii and K. pneumoniae significantly contribute to mortality in intensive care units (ICUs), with a mortality rate of over 50% (Balkhair et al., 2019; Brink, 2019) . In this study, we developed and validated a ddPCR-based method to detect A. baumannii and K. pneumoniae in blood samples from patients with suspected BSI. Our results suggest that ddPCR represents a promising method for the accurate and rapid diagnosis of BSIs caused by A. baumannii and K. pneumoniae. A total of 170 patients were recruited from Zhejiang Provincial People's Hospital, Hangzhou, China, from March 2019 to October 2020. The inclusion criteria were age >18 years and suspected BSI. Suspected BSI was defined based on a sudden high fever (temperature ≥38.5°C) accompanied by hemodynamic instability that could not be explained by a site-specific infection at another body site, and an increase of 2 points or more in the sepsis-related organ failure assessment (SOFA) score as previously described (Hu et al., 2021) . The demographic and clinical characteristics of each patient were collected within the first 24 h of suspicion of BSI; details are summarized in Table 1 . Telephone follow-up interviews were conducted with the surviving patients. An unfavorable outcome was defined as 28-day all-cause mortality after ICU admission. Upon preliminary confirmation of BSI, whole blood samples were obtained for blood culture and molecular diagnosis. Two sets of blood cultures were collected from each patient according to routine clinical practice; each set consisted of one aerobic bottle and one anaerobic bottle. The blood cultures were incubated at 37°C in a BacT/ALERT ® 3D System (BioMérieux). When a positive signal was obtained, Gram staining was performed, followed by subculturing on a Columbia blood agar plate at 37°C with 5% CO 2 . Following overnight incubation, the pathogens were further identified via matrix-assisted laser desorption-ionization time-of-flight mass spectrometry (VITEK ® MS system; BioMérieux For control bacterial strains, total DNA was isolated from 1 ml of overnight bacterial culture using a TIANamp bacteria DNA kit (TIANGEN Biotech) following the manufacturer's instructions. For blood samples, plasma was obtained via centrifugation at 1600 × g for 20 min. The DNA was extracted from 2 ml of plasma containing 1 μl of internal control using a magnetic serum/plasma DNA kit (TIANGEN Biotech) and an Auto-Pure20B nucleic acid purification system (Hangzhou Allsheng Instruments). The DNA was eluted in 50 μl of elution buffer and stored at −80°C until use. Primers and TaqMan MGB probes (Table 3) were designed using Primer Express (Thermo Fisher Scientific) and synthesized by General Biosystems. Either a carboxyrhodamine or Cy5 dye reporter was incorporated at the 5′-end of each probe and a nonfluorescent quencher was incorporated at the 3′-end. The sensitivity and specificity were evaluated by performing sequence alignments using GenBank data and the Basic Local Alignment Search Tool in NCBI, respectively. Analytical specificity was evaluated by detecting the genomic DNA isolated from the positive and negative control bacterial strains. Analytical sensitivity was determined via limit of blank (LoB) and limit of detection (LoD) assays using 22 replicate samples. Probit analysis was performed to measure the LoD of each bacterium. The mean copy numbers and standard deviations (SDs) were calculated. LoB was calculated as mean blank + 1.645 (SD blank ). LoD was calculated as mean blank + 3 (SD blank ) (Armbruster & Pry, 2008 Acinetobacter baumannii and Klebsiella pneumoniae detection performed to analyze non-normally distributed continuous variables. Categorical variables were reported as frequencies and percentages and were analyzed using the chi-squared test. Values of p < 0.05 were considered significant. We designed a specific primer-probe set for each strain to detect A. baumannii and K. pneumoniae in the blood samples. The specificity test showed that all 5 A. baumannii isolates and the 33 K. pneumoniae isolates from ATCC were detected using the corresponding primer-probe set, whereas the 131 negative control isolates were not detected using the primer-probe sets (Table 4 ). These results suggest that each primer-probe set is specific to either A. baumannii or K. pneumoniae. Analytical sensitivity was measured using the LoB and LoD assays. The LoB of A. baumannii and K. pneumoniae were 0.55 copies/µl and 0.15 copies/µl, respectively; whereas the LoD of A. baumannii and K. pneumonia were 0.93 copies/µl and 0.27 copies/µl, respectively (Table 4 ). The LoD was defined as the lowest concentration that yielded positive results in the ddPCR assay. Within-run precision (repeatability) and within-laboratory precision (reproducibility) were measured using different concentrations of sheared A. baumannii and K. pneumoniae DNA samples. The withinrun coefficients of variation (CVs) of A. baumannii and K. pneumoniae were 2.7%-5.3% and 2.3%-9.5%, respectively. The within-laboratory CVs of A. baumannii and K. pneumoniae were 5.3%-7.5% and 7.4%-9.7%, respectively (Tables 5 and 6 ). Linearity was determined using two-fold serial dilutions of the DNA template. The regression lines representing the linear relationships between DNA copy numbers and concentrations of A. baumannii (R 2 = 0.9925) and K. pneumoniae (R 2 = 0.9915) are shown in Figure 1 . Taken together, these results indicate that ddPCR has excellent repeatability, reproducibility, and linearity in detecting A. baumannii and K. pneumoniae in blood samples. Blood samples were obtained and transported to the ddPCR laboratory in approximately 10 min. Plasma was immediately isolated after centrifugation for 20 min. DNA extraction and PCR amplification were completed within 3 h. Data analysis was performed within 30 min using the GenePMS software. The average turnaround time for the ddPCR assay was 4.2 ± 0.51 h, which was remarkably shorter than that of the blood culture method (90.6 ± 10.9 h, p < 0.01). 6.5%, 11/170) ( Table 7) . The clinical characteristics of the 53 patients that tested positive via ddPCR assays are summarized in Table 3 . No significant differences were observed in age, systolic and diastolic blood pressure, plasma C-reactive protein levels, white blood cell counts, serum creatinine, or the use of vasoactive drugs (all p ≥ 0.05) between the patients that tested positive based on ddPCR alone (n = 42) and the patients that tested positive via both ddPCR and blood culture (n = 11). Compared to 11 patients that were reported positive via (0) Abbreviations: LoB, limit of blank; LoD, limit of detection; SD, standard deviation. Analytical specificity and sensitivity of droplet digital polymerase chain reaction for Acinetobacter baumannii and Klebsiella pneumoniae detection non-severe BSIs that may be misdiagnosed via blood culture assays, may still benefit from ddPCR testing and achieve better clinical outcomes. Blood culture is performed as the gold standard for detecting pathogens in BSIs; however, it has a long turnaround time and relatively low sensitivity. By studying 165,593 blood specimens from 13 hospitals in the USA, Tabak et al. showed that the median time required to identify BSI pathogens using traditional blood culture is 44.0 h, with a sensitivity of approximately 70% for critically ill patients and even lower for fastidious microorganisms (Tabak et al., 2018) . To overcome the shortcomings of blood culture in BSI diagnosis, we developed a culture-independent ddPCR method to rapidly and accurately identify A. baumannii and K. pneumoniae in blood samples of patients with suspected BSIs. Since ddPCR shows higher sensitivity, accuracy, and replication than those of conventional methods, it has a great potential to identify pathogenic microorganisms. In addition to the use of broad-range primers for the amplification of highly conserved bacterial 16S rRNA and fungal 28S rRNA which enables the detection of BSIs and discrimination between fungal and bacterial infections (Wouters et al., 2020; Ziegler et al., 2019) , ddPCR is also widely applied to detect and quantify specific pathogenic microorganisms, such as Hepatitis B virus (HBV) (Huang et al., 2015) , Mycobacterium tuberculosis (Devonshire et al., 2015; Pan et al., 2021) , Aspergillus terreus (Alanio et al., 2016) , methicillin-resistant Staphylococcus aureus (Luo et al., 2017) , and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) Suo et al., 2020) . Recently, in the application of SARS-CoV-2 nucleic acid detection, ddPCR has been shown to offer greater advantages for the clinical diagnosis of Coronavirus Disease 2019 (COVID-19) to reduce false-negative results compared to those obtained via quantitative reverse transcription-PCR (RT-qPCR) , although RT-qPCR is widely used as the gold standard for the clinical detection of SARS-CoV-2. In discharged patients with COVID-19, ddPCR is more suitable for quantification of SARS-CoV-2 with a lower viral load compared to using RT-qPCR, and significantly improves the accuracy of diagnosis of patients with COVID-19 who relapse . Our results showed that ddPCR could identify A. baumannii and K. pneumoniae in whole blood samples within 4 h, with a specificity of 100% for each strain and detection limit of 0.93 copies/µl for A. baumannii and 0.27 copies/µl for K. pneumoniae. This is consistent with the results of a previous study which reported that ddPCR requires only 0.54 ± 0.94 copies of covalently closed circular DNA for accurate HBV detection (Mu et al., 2015) . Clinical validation of 170 patients with suspected BSI showed that ddPCR not only identified patients who tested positive based on blood culture but also detected those who tested negative via blood culture. Notably, compared to patients who tested positive via both blood culture and Accumulating evidence has demonstrated the feasibility of NGS of plasma cfDNA to identify pathogens in BSI (Blauwkamp et al., 2019; Grumaz et al., 2019; Rossoff et al., 2019) . However, the typical turnaround time of 2 d and the high cost of NGS represent barriers to the application of cfDNA NGS in clinical practice (Crawford et al., 2020; Simner et al., 2018) . In this study, we exploited the ultra-high sensitivity of ddPCR and the feasibility of cfDNA in pathogen identification to develop a ddPCR-based assay using cfDNA as the template. The turnaround time of ddPCR to diagnosis was estimated as 4 h, which is significantly shorter than that of NGS (2-3 days) (Grumaz et al., 2016) or blood culture (90.6 ± 12.9 h in this study). and fungi using metagenomic DNA as the template and broad-range primer-probe sets; however, the overall specificity in clinical validation is only 80% (Wouters et al., 2020) . In this study, we used cfDNA as the template and designed specific primer-probe sets for A. baumannii and K. pneumoniae. We achieved 100% specificity for detecting well-characterized ATCC and CMCC isolates of each species, which was higher than those reported for other PCRbased methods, such as SeptiFast (50% according to Warhurst et al., 2015, and 85 .5% according to Korber et al., 2017) The sensitivity of blood culture assays is typically low; the blood culture-based positivity rate of patients with sepsis has been reported as 51% over 22 years in the United States (Martin et al., 2003) . Cheng et al. reported a 71.7% blood culture-based positivity rate in patients with severe sepsis from ten university hospitals in China (Cheng et al., 2007) . Similarly, 70% of patients with sepsis admitted to the ICU in a 1-day international investigation were reported to be positive based on blood culture tests (Vincent et al., 2009 ). The low sensitivity of blood cultures may be attributed to the low abundance of bacteria in the blood, antibiotic treatment before sampling, and culture techniques. Molecular detection methods are less affected by these factors; thus, they usually report higher positivity rates than those obtained via blood cultures. The positivity rates determined via different molecular methods are 1.56 to 6.45-fold higher than those determined via blood culture (Farnaes et al., 2019; Grumaz et al., 2019; Korber et al., 2017; Long et al., 2016; Nguyen et al., 2019 In this study, the 53 patients who tested positive via ddPCR assay had typical symptoms resulting from BSIs, including body temperature above 38.5°C, abnormally elevated serum levels of C-reactive protein, and procalcitonin, hemodynamic instability, and severe organ dysfunction. Notably, the patients who tested positive via ddPCR alone exhibited less severe symptoms than those reported positive via both ddPCR and blood culture, suggesting that ddPCR is more sensitive than blood culture for early diagnosis of BSIs. This study had certain limitations. First, as the microbial cfDNA is possibly released from dead pathogens and is continuously detected almost 2 weeks after conventional blood cultures report negative results (Eichenberger et al., 2021) , the presence of microbial cfDNA in the blood does not necessarily indicate the presence of living microorganisms in the bloodstream and may also be derived from a previous BSI. Second, unlike metagenomic NGS that has a wider coverage for detecting causative pathogens, ddPCR can only identify a small number of target pathogens; thus, it is not possible to determine whether a patient has a mono-microbial infection or polymicrobial infection based on the detection results from the ddPCR assay alone. In addition, for multiple pathogens detected via ddPCR, it is difficult to determine if a pathogen associated with a relatively low cfDNA load in the blood sample is clinically relevant. Third, it is challenging to distinguish causative pathogens from normal microbes or environmental contaminants. Although the ddPCR assay can report quantitative results, there are no clear cutoffs that differentiate infection from colonization or contaminants. Therefore, to partially overcome the aforementioned limitations, enlargement of assay panels to cover additional pathogens and tracking the microbial cfDNA load change may be helpful, as the dynamic monitoring of microbial cfDNA concentration may reflect the progress of active BSIs. When determining the clinical significance of a pathogen detected via cfDNA sequencing, the entire clinical scenario should be considered. In conclusion, we developed a novel ddPCR method to detect two major pathogens in patients with suspected BSI. Clinical validation revealed that our method was superior to blood culture in terms of specificity, sensitivity, and turnaround time, and represents a promising method for the early and accurate diagnosis of BSIs. However, in this pilot study, we only evaluated two major Gramnegative bacteria responsible for BSIs. Other clinically important pathogens should be investigated in future studies. The study was financially supported by grants from the Key Research and Development Project of the Science Technology Department of Zhejiang Province (2020C03031), the National Natural Science Foundation of China (81971857), and the Natural Science Foundation of Zhejiang Province (LY17H150005). None declared. The study protocol was approved by the Institutional Review Board and Ethics Committee of Zhejiang Provincial People's Hospital (No. 2019KY002). Written informed consent was obtained from all patients or their legal representatives. All data generated or analyzed during this study are included in this published article. https://orcid.org/0000-0001-6246-5258 Variation in copy number of the 28S rDNA of Aspergillus fumigatus measured by droplet digital PCR and analog quantitative real-time PCR Next-generation sequencing of circulating tumor DNA for early cancer detection Limit of blank, limit of detection and limit of quantitation Prevalence and 30-day all-cause mortality of carbapenem-and colistin-resistant bacteraemia caused by Acinetobacter baumannii, Pseudomonas aeruginosa, and Klebsiella pneumoniae: Description of a decadelong trend Analytical and clinical validation of a microbial cell-free DNA sequencing test for infectious disease Robert Koch and the 'golden age' of bacteriology … Circulating Donor-Derived Cell-Free DNA in Blood for Diagnosing Active Rejection in Kidney Transplant Recipients (DART) Study Investigators Epidemiology of carbapenem-resistant Gramnegative infections globally Effectiveness of practices to increase timeliness of providing targeted therapy for inpatients with bloodstream infections: a laboratory medicine best practices systematic review and meta-analysis Droplet digital PCR as an emerging tool in detecting pathogens nucleic acids in infectious diseases Epidemiology of severe sepsis in critically ill surgical patients in ten university hospitals in China* Investigating transfusion-related sepsis using culture-independent metagenomic sequencing Highly reproducible absolute quantification of Mycobacterium tuberculosis complex by digital PCR Microbial Cell-Free DNA Identifies Etiology of Bloodstream Infections, Persists Longer Than Conventional Blood Cultures, and Its Duration of Detection Is Associated With Metastatic Infection in Patients With Staphylococcus aureus and Gram-Negative Bacteremia Noninvasive diagnosis of fetal aneuploidy by shotgun sequencing DNA from maternal blood Communityacquired pneumonia in children: cell-free plasma sequencing for diagnosis and management Benefit of appropriate empirical antibiotic treatment: Thirty-day mortality and duration of hospital stay Enhanced performance of next-generation sequencing diagnostics compared with standard of care microbiological diagnostics in patients suffering from septic shock Next-generation sequencing diagnostics of bacteremia in septic patients Liquid biopsy for infectious diseases: A focus on microbial cell-free DNA sequencing Comparison of blood pathogen detection among droplet digital PCR, metagenomic next-generation sequencing, and blood culture in Critically Ill patients with suspected bloodstream infections Next generation digital PCR measurement of hepatitis B virus copy number in formalin-fixed paraffinembedded hepatocellular carcinoma tissue Considerations for digital PCR as an accurate molecular diagnostic tool Detection of methicillin-resistant staphylococcus aureus by a duplex droplet digital PCR assay SeptiFast versus blood culture in clinical routine -A report on 3 years experience Duration of hypotension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock Microbial cell-free DNA in plasma of patients with sepsis: a potential diagnostic methodology Evaluation of droplet digital PCR for quantification of SARS-CoV-2 Virus in discharged COVID-19 patients Accurate detection of methicillin-resistant staphylococcus aureus in mixtures by use of single-bacterium duplex droplet digital PCR The T2Bacteria Panel had 90% sensitivity for detecting targeted organisms, 43% for any bloodstream infection organism The epidemiology of sepsis in the United States from Long-term morbidity and mortality following bloodstream infection: A systematic literature review A sensitive and accurate quantification method for the detection of hepatitis B virus covalently closed circular DNA by the application of a droplet digital polymerase chain reaction amplification system Performance of the T2Bacteria panel for diagnosing bloodstream infections: A diagnostic accuracy study Mycobacterium tuberculosis-derived circulating cell-free DNA in patients with pulmonary tuberculosis and persons with latent tuberculosis infection Comparison of digital PCR and quantitative PCR with various SARS-CoV-2 primerprobe sets Digital PCR to detect and quantify heteroresistance in drug resistant mycobacterium tuberculosis Droplet-based digital PCR and next generation sequencing for monitoring circulating tumor DNA: A cancer diagnostic perspective Early identification and treatment of pathogens in sepsis: Molecular diagnostics and antibiotic choice Noninvasive diagnosis of infection using plasma next-generation sequencing: A single-center experience Clinical utility of droplet digital PCR for human cytomegalovirus Identification of chromosomally integrated human herpesvirus 6 by droplet digital PCR Understanding the promises and hurdles of metagenomic next-generation sequencing as a diagnostic tool for infectious diseases ddPCR: a more accurate tool for SARS-CoV-2 detection in low viral load specimens Blood culture turnaround time in U.S. acute care hospitals and implications for laboratory process optimization A multiplex droplet digital PCR assay for non-invasive prenatal testing of fetal aneuploidies International study of the prevalence and outcomes of infection in intensive care units Diagnostic accuracy of SeptiFast multi-pathogen real-time PCR in the setting of suspected healthcare-associated bloodstream infection Droplet digital polymerase chain reaction for rapid broad-spectrum detection of bloodstream infections Evaluation of the magicplex sepsis real-time test for the rapid diagnosis of bloodstream infections in adults Detection of pathogenic microorganisms from bloodstream infection specimens using TaqMan array card technology 16S rDNA droplet digital PCR for monitoring bacterial DNAemia in bloodstream infections How to cite this article: Zheng, Y., Jin, J., Shao, Z., Liu, J., Zhang, R., Sun, R., & Hu, B. (2021)