1 Introduction

The ongoing COVID-19 pandemic has had devastating effects on populations, social structures, and economic growth. These are further exacerbated by the increasing extent of global connectivity and geographical mobility, which expedite infection spread at an uncontrollable pace. The causative agent of this outbreak has been identified as the recently discovered SARS-CoV-2 coronavirus. Effective control and containment of this pathogen require reliable diagnostic assays and potent therapeutic agents. Recent advances in computational technology and biomedical engineering have placed a toolkit of immense potential in our hands, with a formerly unimaginable capability to enable multidisciplinary innovations and highly accelerated discoveries. In this review, we provide a critical evaluation of diagnostic techniques and treatment strategies targeted at the SARS-CoV-2. Alongside an eclectic selection of such reports, our analysis also focuses on the applications, advantages, and pitfalls of emerging technologies like CRISPR/Cas, immune-informatics, artificial intelligence, and machine learning.

2 Diagnostic tools for detecting SARS-CoV-2

Figure 1 provides a graphical summary of the various approaches discussed herein.

Figure 1
figure 1

Summary of the various approaches towards diagnosis of SARS CoV-2 infection.

Full size image

2.1 RT-PCR

PCR-based methods are considered the gold standard for viral detection. SARS-CoV-2 requires RT-PCR-based approaches, by virtue of being an RNA virus. In this section, we highlight some interesting developments as well as possible pitfalls of this technique, in the context of viral diagnosis.

One of the earliest workflows set up in response to the outbreak was by Christian Dorsten and colleagues (Corman et al. 2020). The team established an accurate, sensitive, and specific RT-PCR protocol against the SARS-CoV-2. Later research, including a comparative study by Nalla et al., have attested to its superior sensitivity (Nalla et al. 2020). A notable feature of this pioneering work was that the assays were designed without access to any actual SARS-CoV-2 genomic specimens or patient samples, by relying on genome information sourced from Chinese researchers and synthetic nucleic acid technology. An interesting comparison between two commercially available testing kits – (the TaqMan™ 2019-nCoV Assay Kit v1 (ThermoFisher) and the 2019-nCoV CDC qPCR Probe Assay (Integrated DNA Technologies)) – was recently reported, results from which indicate that the former is capable of reliably detecting SARS-CoV-2 presence in nasopharyngeal swab samples without any RNA extraction steps (Beltrán-Pavez et al. 2020). Most diagnostic devices relying on RT-PCR-based amplification suffer from the drawback of requiring either sample lysis or purified nucleic acid samples, a step that involves additional reagents, increased testing time, human errors and costs, in addition to significant compromises on the field-applicability of the test. Subject to further validation, the above result represents a potential breakthrough for rapid and accessible diagnosis in low-resource settings, by enabling the development of accessible point-of-care devices. Additionally, while nasopharyngeal/oropharyngeal swabs are recommended sample collection techniques for SARS-CoV-2 testing, these pose a high risk of exposing healthcare workers to large numbers of potentially infected individuals, in addition to causing patient discomfort. An alternative in the form of saliva specimens as a non-invasive method of sample collection has been proposed to address these concerns (To et al. 2020).

In order to enhance the sensitivity and fidelity of RT-PCR assays, a worthwhile strategy is to evaluate the potential of various viral components as probe targets and compare the relative performance of each. For instance, an early RT-PCR-based detection assay used probes against the ORF1b and N region of the viral genome (Chu et al. 2020). Subsequently, another study developed three novel assays aimed at the viral spike protein (S), nucleocapsid (N), and RNA-dependant RNA polymerase (RdRp)/Helicase (Hel) genes, out of which, the RdRp/Hel-based assay demonstrated no cross-reactivity with common respiratory pathogens as well as the lowest detection limit in vitro (Chan et al. 2020). Exploring such novel targets may help develop superior assays, which can improve the sensitivity and accuracy of detection. Furthermore, a comparative analysis by Nalla et al. evaluated seven different assay kits and reported the E-gene primer/probe set (as described by Corman et al. (2020)) and the N2 set (developed by the Division of Viral Diseases, Center for Disease Control and Prevention) as the most sensitive assays available. The importance of such independent assessments of reported/published studies needs to be underscored in healthcare crisis scenarios such as the present one, wherein resources are limited and must be judiciously allocated.

2.2 Chest CT scans

A few studies have also pointed out the liabilities of RT-PCR-based techniques. While some publications and in vitro assays may report excellent parameters of performance, clinical testing and feedback raise several red flags. A most elementary instance of this was reported in a case wherein two patients who were SARS-CoV-2 infected returned negative RT-PCR results, leading the authors to suggest chest CT scans as an essential part of clinical diagnosis (Li et al. 2020a). Similar issues with a high false-negative rate of RT-PCR testing on patients over the course of disease progression and treatment have been reported separately as well (Li et al. 2020e). Such observations have promoted recommendations to include clinical parameters such as CT scan diagnosis as factors to be considered in addition to the aforementioned assays while making decisions on patient discharge, evaluation of recovery/response to treatment protocols. A retrospective study reached similar conclusions in favor of CT scans, ascribing a 97.2% sensitivity to CT, as compared to 83.3% using RT-PCR (Long et al. 2020a).

At this juncture, we would like to emphasize a few recent studies which have underscored the potential of chest CT scans as a viable diagnostic technique. Work by Cao et al. and Dai et al. seeks to formulate a standard set of characteristic features indicative of a SARS-CoV-2 infection. The former, a systematic review and meta-analysis of infected patients, establishes the most common clinical symptoms as fever, cough, chest distress, fatigue, and dyspnea, along with major imaging features as ground-glass opacities and bilateral pneumonia (Cao et al. 2020a). The latter provides a more comprehensive treatment in which over two hundred patients were studied to report bilateral multiple lung lobes in both the periphery and lower portion of the lungs as being present in over 94.98% of the cases (Dai et al. 2020). Additionally, ground-glass opacities and vascular enhancement were identified as characteristic signatures, as well as fibrosis, air, trapping, and interlobular septal thickening. It is worthwhile to note that highly concordant observations have been made across different patient cohorts, and the general agreement between these is an indicator of the proposed technique’s versatility and broad-spectrum applicability. A more extensive and detailed elucidation of the same may be found in the recent review by Hani et al. (Hani et al. 2020). Furthermore, time-point-based CT scan data may be an invaluable tool to assess disease progression, treatment response, development of complications, and understanding the pathophysiology of the disease, as has been stressed by Bernheim et al. (2020). Additionally, an artificial intelligence (AI)-based platform has also been devised which utilizes a neural network to evaluate X-ray images and automatically detects SARS-CoV-2-infected positive specimens with an accuracy of 97.82% (Apostolopoulos and Bessiana 2020). Harnessing advances in computer science for medical diagnostics may serve as an essential tool to standardize and speed up the testing process. Similar applications of machine learning and AI to aid in the pandemic control efforts will be further elaborated in Sect. 4.

2.3 Antibody-based techniques

The evident advantage of antibodies over nucleic acids amplification is higher specificity, accuracy, as well as rapid testing times since it does not require time-consuming amplification steps, nor involves cumbersome extraction/purification processes. A recent report has also highlighted that antibody titers may be a more reliable indicator of SARS-CoV-2 infections, especially for asymptomatic and suspected carriers who return negative RT-PCR results (Long et al. 2020b).

The classic Enzyme-linked Immunosorbent Assay (ELISA) may be readily adapted to serve as a SARS-CoV-2 detection test. Accordingly, this has been employed against the recombinant viral nucleocapsid proteins and the recombinant viral spike proteins, which were used to detect the IgM and IgG antibody levels (Liu et al. 2020). This study established that spike protein-based assay would be more sensitive for detecting IgM. Along similar lines, the comparative performance of ELISA-based and Gold-immunochromatographic Assay (GICA)-based methods have also been evaluated (Xiang et al. 2020). The rapid testing times and high sensitivity observed posit serological assays as superior alternatives to RT-PCR on account of ease of sample collection (blood, which poses minimal risk to healthcare staff compared to swab samples) and low false-negative rates. In this regard, an efficient point-of-care device for diagnostic testing which combined the IgG-IgM testing platform with a testing time of fifteen minutes was recently reported (Li et al. 2020f). While the repurposing of cross-reactive antibodies from highly similar species like the SARS-CoV is a viable idea, an interesting report notes some pitfalls (Tian et al. 2020). Despite the relative identity between the SARS-CoV and the SARS-CoV-2, the authors point out that highly potent receptor binding site-targeting SARS-CoV neutralizing antibodies (like the CR3014 and m396) fail to engage the SARS-CoV-2 spike protein. This indicates that effective management of novel pathogens may necessitate the development of novel antibodies and those homology-based strategies to repurpose existing therapeutics should be rigorously vetted.

However, an over-reliance on antibody-based approaches is impractical. Designing nucleic acid probes is a far more straightforward process, and one much more rapidly deployable. Also, the detection of antibodies may be difficult in the early stages of infection, until a significant titer of IgG/IgM has accumulated in circulation. This may contribute towards significant false-negative rates during large-scale screenings, especially when periodic sampling and extended observation of test subjects may not be feasible. Hence, despite the salient advantages of antibody-based techniques as discussed above, several assay kits and field-testing protocols have remained in the domain of PCR amplification-based methods. Accordingly, our object of focus for the next section will be one such recently evolved technology, loop-mediated isothermal amplification (LAMP).

2.4 RT-LAMP

LAMP-based protocols enable the efficient amplification of nucleic acids at a single point temperature. This feature makes it a strong contender for direct field applications, since incorporating the thermal cycling steps in PCR assays has traditionally been a significant limitation for point-of-care devices. LAMP’s inherent advantages are not limited to isothermal operability alone. This robust technique works efficiently even with crude sample preparations, compared to traditional PCR methods, and also offers a very high amplification efficiency since it is not limited by a doubling-per-cycle threshold (Becherer et al. 2020).

We wish to highlight two particular tests, both of which report colorimetric detection capability – a major boon for point-of-care devices. A team led by Di Liu and Jing Yuan reported RT-LAMP assays for SARS-CoV-2 detection, with ORF1ab and S genes as the primer-probe targets (Yan et al. 2020). The work claims complete detection within 60 min, using a colorimetric detection system that employs fluorescent calcein, where a color change from orange to green indicates positive reactions (visually detectable by the naked eye). Another effort by Renfei Lu et al. involving primers targeted at the RdRp utilizes cresol red (a pH-sensitive indicator dye) for the assay readout (Lu et al. 2020). Since a proceeding amplification can progressively change the buffer pH from alkaline to acidic, a color change from burgundy to orange/yellow signals a positive reaction. These innovations in assay readout methodology imply a significant advancement since they make it possible to have even untrained healthcare workers administer, conduct, and interpret diagnostic tests.

A noteworthy instance of adapting RT-LAMP for a truly bedside point-of-care application is the innovative closed-tube test developed by Song and colleagues (El-Tholoth et al. 2020). Combining straightforward sample collection with single or two-step RT-LAMP amplification protocols, along with a visual detection system based on LCV (leucocrystalviolet – an intercalating agent which colorimetrically detects double-stranded LAMP amplicons), this device is a readily deployable and highly portable testing method that promises to be a cheap and reliable alternative suitable for all testing environments.

2.5 Emerging techniques: reactive polymers and CRISPR/Cas-based systems

This subsection is devoted to a couple of non-conventional detection methods, which may well serve as a versatile platform in the future. The first is a reactive polymer-grafter device which utilizes antibodies to detect dsRNA (Ku et al. 2020). A suitable polymer-coated surface-reactive poly(pentafluorophenyl acetate) in this case is used to immobilize J2 antibodies on the surface. Following the preparation of the platform, it may be directly applied to detect dsRNA as the J2 antibodies may bind dsRNA molecules specifically. The principal advantage, in this case, is that the antibody binding is sequence-independent; hence the platform may serve as a universal virus detection unit. However, it may be noted that such one-size-fits-all approaches suffer from an inherent lack of specificity. In this case, the sequence-independent nature of detection suggests a high possibility of false positives. Such a scenario would necessitate secondary tests using established protocols for reliable confirmations of diagnosis, thereby positing the J antibodies-based technique as a potential preliminary screening step, rather than as a standalone entity.

The second major innovation involves the use of CRISPR-Cas-based techniques for the detection of viral nucleic acids. Recently, such CRISPR-based detection systems, notably SHERLOCK and DETECTR, were reported as emerging diagnostic tools. The fundamental principle involves the activation of a suitable Cas variant (Cas13 for SHERLOCK, Cas12a for DETECTR) by binding of the appropriate target sequence (ssRNA or mRNA for Cas13 and dsDNA for Cas12a), which enables the Cas variants to promiscuously cleave and degrade surrounding ssRNA and ssDNA, respectively. The ssRNA/ssDNA meant for the latter reaction may be coupled with a quenchable fluorophore, which produces a quantifiable signal upon being released by cleavage. Hence, the detection of the exact target sequence is indicated by an assay readout in the form of a fluorescence signal. This system allows for rapid and accurate detection of viral RNA samples since amplification of these (along with suitable modifications as per the system used) can then allow for precise targeting and detection by the Cas-based assay described above. This remarkable ability has been recently demonstrated by using the DETECTR system as a diagnostic tool for SARS-CoV-2 infections in a rapid lateral flow assay incorporating RT-LAMP using nasal swab samples (Broughton et al. 2020). Additionally, a series of reports may be found on the Broad Institute’s website (https://www.broadinstitute.org/news/enabling-coronavirus-detection-using-crispr-cas13-open-access-sherlock-research-protocols-and), wherein a team comprising of Feng Zhang, Omar Abudayyeh, and Jonathan Gootenberg aim to develop paper-based platforms for rapid and straightforward SARS-CoV-2 detection (B Institute 2020).

3 Treatment strategies for SARS-CoV-2 infections

In this section, we endeavor to review some of the established and emerging treatment paradigms which are being currently employed or recommended on the basis of published literature and empirical evidence from clinical case studies. The major ones have been pictorially represented in figure 2.

Figure 2
figure 2

Some established and emerging treatment paradigms being currently employed or recommended on the basis of published literature and empirical evidence from clinical case studies.

Full size image

3.1 Drugs-based conventional approaches

Amongst drug-based treatment strategies, the broad-spectrum antiviral chloroquine and its derivative hydroxychloroquine have received much attention. As early as February 2020, both remdesivir and chloroquine were identified as potential inhibitors of the SARS-CoV-2 in vitro (Wang et al. 2020b). Remdesivir, a potential candidate against Ebola, SARS-CoV, and MERS-CoV, is an adenosine analog that inhibits viral replication. On the other hand, chloroquine targets endosomal fusion by elevating the endosomal pH, alongside interfering with the glycosylation of ACE2. The rationale for using chloroquine is further buttressed by a variety of possible antiviral effector functions (Devaux et al. 2020). Experimental evidence for a putative mechanism was provided by structural data and molecular modeling approaches, where a conserved ganglioside binding domain at the N terminus of the viral S protein was identified as the target for chloroquine and hydroxychloroquine. This inhibition of viral attachment and entry appears to explain chloroquine and its derivative hydroxychloroquine’s apparent efficacy against SARS-CoV-2 infections (Fantini et al. 2020). However, it is important to recognize that evidence for the clinical benefits of these much-touted drugs is contentious at best and unreliable at worst. While hydroxychloroquine and azithromycin were earlier claimed as potentially effective drugs against SARS-CoV-2 infections, the reporting study has since come under criticism for the lack of a control group and incomplete descriptors of actual clinical outcomes and patient inclusion criteria (Gautret et al. 2020). A separate randomized trial enrolling 62 patients did, however, report a statistically significant positive effect of hydroxychloroquine treatment in alleviating patient symptoms (Chen et al. 2020c). These claims are offset by three additional reports, none of which conclude any significant benefit from this treatment method (Borba et al. 2020; Magagnoli et al. 2020; Mahevas et al. 2020). An emerging class of candidate drugs are corticosteroids, which have recently generated great interest in the context of COVID-19 treatment approaches. While early skepticism did reign over corticosteroids usage, considering its propensity to elicit adverse reactions in severely ill patients, evidence from recent clinical trials and meta-analyses have hinted at significant benefits from this drug for the management of critically ill patients. This has led to a growing advocacy of corticosteroids as a leading candidate in the global effort against COVID-19 (Halpin et al. 2020; Ma et al. 2020b; Sanders et al. 2020; Singh et al. 2020; Yang et al. 2020b; Ye et al. 2020; Zha et al. 2020).

Emerging technologies powered by AI offer systematic and highly modular solutions for urgent drug discovery and identification requirements. A study published in The Lancet (Richardson et al. 2020) employed Benevolent AI, a platform for AI-guided drug discovery and clinical development, to offer suggestions for drugs targeted explicitly at SARS-CoV-2. Using a systematic knowledge graph, the platform was able to identify baricitinib, a Janus kinase inhibitor that binds the AAK1 (AP2-associated protein kinase 1, a regulator of endocytosis). Section 4 will deal with this domain in greater detail.

3.2 Docking simulations and molecular dynamics-based approaches

Powerful computational resources can greatly enhance our ability to rapidly screen for drugs and inhibitory agents against emerging pathogens. The SARS-CoV-2 protease has been a popular target for such docking simulations and molecular dynamics-based studies (Liu and Wang 2020), in part due to the recent availability of its structure. A broad screening effort of over 1.3 billion compounds from the ZINC15 database used an upgraded docking protocol (termed DeepDocking) to identify over a thousand such potential inhibitors of the SARS-CoV-2 protease (Ton et al. 2020). A more systematic approach along similar lines screened 687 million compounds via docking, followed by molecular dynamics simulations to evaluate ligand-binding energies, stability, toxicity assessment, as well as off-target binding (Fischer et al. 2020). Alongside such screening efforts for existing molecules, the identification of promising lead compounds that could serve as scaffolds for further modification is an essential step to enable the development of novel drugs. This requirement was addressed by a recent in-silico study, which provided a clustered report of twenty short-listed compounds into three groups, each of which could serve as leads for the development of SARS-CoV-2 inhibiting drugs (Ortega et al. 2020).

While the ability of advanced computational facilities to virtually screen multitudinous drugs and compounds – numbering anywhere from a few thousand to even a billion – might be unmatched by traditional methods, the importance of physical assay-based screens is indispensable. The recent report of a large-scale (12,000+ compounds) repositioning survey of FDA-approved and under-trial drugs in the context of SARS-CoV-2-targeted antivirals attests to this (Riva et al. 2020). Following an initial selection based on inhibition of cytopathic effects, an orthogonal validation assay was used to delineate the compounds which specifically inhibited viral replication. This study also provided classification and categorization of target annotations and drug target genes for the short-listed drugs. The identification of these specific domains can help similar such efforts in the future to be more focussed on specific drug types and targets that have been identified to exert the most promising antiviral effect on SARS-CoV-2. Another notable effort combined virtual screening methods with high-throughput assays to enable the rapid identification of promising drug candidates (Jin et al. 2020). Using a FRET-based assay to assess the enzymatic activity of SARS-CoV-2 protease, over 10,000 compounds were screened for inhibitory effects on the enzyme. Additionally, the authors extensively analyzed a compound termed N3, initially developed using computer-aided drug design as an inhibitor of multiple coronavirus proteases. Following molecular docking, kinetic analysis of the protease, and determination of its crystal structure with the SARS-CoV-2 protease, the inhibitory effects of N3 were ascertained. The comprehensive approach adopted by this study is an exemplar model for similar investigations. The obvious merits of virtual screening methods notwithstanding, strong emphasis needs to be laid on developing high-throughput biochemical and cellular assays for experimental validation of pharmacologically active compounds. The importance of this cannot be overstated, as such empirical evidence helps establish a reliable standard for ascertaining the actual merits of proposed therapeutic approaches, which may, at least at the current stage of research, only be supplemented by computational studies.

3.3 Plasma therapy

Alongside treatment with drugs, convalescent plasma (CP) therapy is being advocated as a clinically effective practice. The procedure involves using blood plasma obtained from recovered patients as an adaptive immune therapy for critical cases. The underlying principle at work is that the donor plasma contains a high titer of antibodies that are effective against the pathogen, hence provides an immunity boost to the recipient. CP has been advocated as an alternative therapy for SARS-CoV-2 infections (Chen et al. 2020a) and therapeutic plasma exchange has been recommended as a suitable strategy to deal with the often associated acute respiratory distress syndrome (ARDS) (Keith et al. 2020). An initial four-patient sample established that CP could have therapeutic benefits for critically ill patients (Zhang et al. 2020a). While all four patients showed a progression towards negative RT-PCR results, the authors of the study stress that it would be difficult to rule out confounding factors as well as the effects of additional supportive care and treatments such as antiviral drugs and intravenous immunoglobulin. A similar study involving a ten-patient cohort demonstrated that CP dosages could either increase or maintain the existing levels of neutralizing antibodies in the recipients (Duan et al. 2020). Additionally, the clinical symptoms were also observed to improve progressively, leading up to the absence of viremia. The study highlights two other salient outcomes – one, that CP appeared to be tolerated well by the recipients without adverse reactions, and two, that transfusion of CP alleviated inflammation. It is important to note, however, that these studies need to be further supplemented by rigorous and controlled clinical trials to discount both the effects of external factors, as well as establish proper protocols for CP therapy concerning the dosage, duration, and other appropriate parameters.

3.4 Managing cytokine storm responses

Among the several complications associated with SARS-CoV-2 infections, cytokine storm responses, which are a by-product of immune hyperactivation and loss of regulatory control over pro-inflammatory cytokines, pose a major risk for patient survival and may lead to rapid deterioration of patient health. The stimulation of CD-4+ T-cells to differentiate into Th1 helper cells along with IL-6, interferon-gamma, and GM-CSF (granulocyte-macrophage colony-stimulating factor) production, amongst other proinflammatory cytokines, are considered as the steps leading up to a cytokine storm-like condition. The persistently high levels of certain cytokines like IP-10, MCP-3, and IL-1ra may be characteristic of such an onset, and these may be developed as screening parameters to ascertain the same (Yang et al. 2020a).

Extensive efforts to safely manage this condition have been undertaken. For instance, the use of an IL-6 receptor antagonist drug such as tocilizumab has been suggested, since the IL-6 has been identified as a key driver of cytokine storm response (Zhang et al. 2020b). Writing in The Lancet, Mehta et al. also recommend similar IL-6 targeted strategies (Mehta et al. 2020). While a previous study had failed to establish a significant connection between the administration of corticosteroids and improved treatment outcomes (Zha et al. 2020), Mehta et al. recommend consideration of such a course of treatment to tackle hyperinflammation. However, there is no concrete evidence to support corticosteroid-based treatment as being of overall therapeutic value, in addition to the associated risks of lung injury (Russell et al. 2020). A detailed review also suggests several strategies by which cytokine storm responses may be dealt with, including the use of the above-cited IL-6R agonist tocilizumab. Glycyrrhetinic acid, another IL-6 and STAT3 signaling inhibitor has been suggested as a potential candidate. Additionally, mesenchymal stem cell infusions have been highlighted due to the immunomodulatory and anti-inflammatory potential of these cells (Cheng et al. 2020). An alternate approach to managing cytokine storm response has been suggested in the form of artificial blood purification systems. Essentially, this involves using an external filtration-cum-treatment setup to modulate the various factors and chemokines responsible for orchestrating the proinflammatory response. This may be achieved by the use of scavenging inflammatory mediators and plasma exchange. At the same time, continuous venovenous hemofiltration units may be employed, as demonstrated for the successful containment of cytokine storm responses in critically afflicted H7N9 influenza patients (Zhang et al. 2020d). The merits of blood purification therapy for alleviating similar symptoms in SARS-CoV-2 infected patients have also been demonstrated by a recent study from Wuhan involving three patients (Ma et al. 2020a).

3.5 Other emerging treatment strategies

Strategies aimed at halting viral entry are not limited to drug-based approaches alone, as peptide inhibitors can play a vital role in achieving similar effects. To this end, the team led by Shibo Jiang and Lu Lu has built upon previous work that generated fusion inhibitors against the SARS-CoV spike protein’s HR1 region to design HR2P as a fusion inhibitor against SARS-CoV-2, as well as EK1 (a pan-coronavirus fusion inhibitor). In a follow-up study, the team generated a series of EK1-derived lipopeptides to determine the ones exhibiting the most potent inhibition of fusion (variant identified as EK1C4). Suitable formulations of such peptide preparations, as suggested by the authors, may be a valuable treatment resource for SARS-CoV-2 infections (Xia et al. 2020a; Xia et al. 2020b).

The value of mesenchymal stem cell transplants in improving patient health for SARS-CoV-2 infections was demonstrated in a clinical study involving seven patients (Leng et al. 2020). Several outcomes were characteristic of the suggestions made by Cheng et al. as mentioned earlier, including immunomodulatory effects, an attenuating effect on proinflammatory cytokines, and elevated IL-10 levels. An interesting observation made by the authors of the clinical study was that such mesenchymal stem cells lacked both ACE2 and TMPRSS2 gene expression, which could effectively render them immune to SARS-CoV-2 infection. The possible benefits of cell therapy-based therapeutic approaches may be immense and have been explored at length in a recent review (Khoury et al. 2020).

The potential of melatonin as an adjuvant in clinical treatment on account of its immune-enhancing, anti-inflammatory, and anti-oxidative effects has been extolled at length in a recent review (Zhang et al. 2020c). Another study focused on the antagonists of the lipid-dependant attachment process of viral particles to host cells (Baglivo et al. 2020). Given that SARS-CoV-2 utilizes structures such as lipid rafts to mediate its entry process, the authors recommend exploring the possibilities of employing substances like cyclodextrin and sterols to interfere with such viral attachment and entry-enabling mechanisms. Along similar lines of conjecturing prospective therapeutic candidates, EZH2-mediated H3K27me3 at ACE2 promoter regions has been demonstrated to inhibit the expression of ACE2 receptors in mammalian cell lines, as indicated by data from RNA-Seq and CHIP-Seq experiments (Li et al. 2020d).

A major cause of patient mortality in severe SARS-CoV-2 infections is the ARDS as well as coagulopathy. Fibrinolytic therapy to target the fibrin deposition in the pulmonary vasculature (a potential contributor towards ARDS) has been shown to improve survival, along with administration of tissue plasminogen activators (which converts plasminogen to plasmin, that can further breakdown blood clots). This strategy has been shown to improve conditions in a three patient study, and further clinical trials to ascertain its applicability as a standard clinical practice for critically ill SARS-CoV-2 infected patients with ARDS have been recommended (Wang et al. 2020a).

The development of effective prophylactic agents is an essential component of a comprehensive approach towards tackling a global pandemic of this scale and severity. The importance of administering prophylaxis to frontline healthcare workers, screening personnel, and researchers, among others, cannot be overstressed, as the safety and continued productivity (which necessitates sound health) of these key players is paramount to maintain any sustained effort at containing the pandemic and providing healthcare and treatment of adequate quality to affected patients. In this regard, a concise review of established protocols and proposed strategies currently under scrutiny has been recently published in the form of correspondence (Agrawal et al. 2020).

Considering the rapidly evolving status of COVID-19 treatment protocols, it is of utmost importance to critically evaluate the most up-to-date evidence available while designing guidelines for clinical practice. Of equal pertinence are social awareness campaigns and the regular dissemination of accurate information regarding clinically proven treatment protocols. This assumes greater importance in light of a growing inclination towards scientifically unsubstantiated claims based on anecdotal/circumstantial evidence or misrepresented facts placed out of context. We further present a brief, non-exhaustive listing (table 1) of leading therapeutic agents frequently associated with COVID-19 treatment, as well as the current opinion of the scientific community regarding each of these, based on the evidence from reported/ongoing clinical trials.

Table 1 Leading therapeutic agents against COVID-19, evaluated and described across common parameters
Full size table

4 Emerging technologies for biomedical research: artificial intelligence and machine learning

The fields of Artificial Intelligence (AI) and Machine Learning (ML) have undergone tremendous growth in scope and popularity in the recent past, establishing them as important paradigms for future innovation. More importantly, the development of modules suitable for biomedical applications has transformed AI and ML into attractive tools for the current global effort to tackle the SARS-CoV-2 pandemic.

With regard to infectious diseases, AI-based platforms may be employed to monitor outbreak and clustering trends, as well as evaluate the impact of containment programs and carry out risk assessment with minimal human intervention. The AI-based service BlueDot deserves a noteworthy mention of having raised one of the earliest alarms on SARS-CoV-2’s potential for broad global spread (Bogoch et al. 2020). Additionally, these hold immense promise for dynamic situational monitoring of patients in real-time, enabling critical evaluation and optimization of trial treatment protocols in a much more efficient manner. ML-based solutions can be expected to consistently enhance and refine such designs over time, provided they are trained on a readily updated and expertly-curated dataset. This represents an important caveat, as the efficacy and accuracy of ML-based pattern prediction and recognition may be maintained only if the training dataset is free of biases and a diverse range of scenarios are well-represented (Fitzpatrick et al. 2020). We had previously referred to the work by Apostolopoulos and Bessiana, which could screen for SARS-CoV-2 infections using X-ray images. This has been supplemented by a neural network-based platform using X-ray images, with a maximum prediction accuracy of 98% (Narin et al. 2020). Furthermore, an automated AI-based protocol to identify SARS-CoV-2 infections and distinguish these from other lung diseases or community-acquired pneumonia has been recently formulated (Li et al. 2020a, b, c, d, e, f). It must be noted that both these models were trained on datasets consisting of actual patient X-ray images, which is in line with a point we have reiterated above – that training datasets should necessarily be accurate, well-represented, and as free of biases as possible. An exciting new development reported by the Broad Institute’s Sabeti lab concerns an ML-based platform for automated Cas13-based SHERLOCK assay design (Metsky et al. 2020). Such an advance represents a potential breakthrough, as highly optimized and expanded open-access platforms of this nature can aid researchers the world over in rapidly designing standardized tests for novel and emerging pathogens using just the knowledge about their genomic sequence. This can help accelerate the deployment of rapid responses to outbreaks and thereby greatly enhance early screening and containment efforts. Understandably, a confident reliance on such automated processes may only be possible once multiple field applications have validated the same and demonstrated similar, if not superior, performance in comparison with existing assay design pipelines.

Section 3.1 of this review highlights the study which determined baricitinib as a potential drug against SARS-CoV-2 (Richardson et al. 2020). Writing in The Lancet, the same team extended this workflow to develop more comprehensive treatment strategies aimed at shortlisting drugs that offered both antiviral as well as anti-inflammatory effects (Stebbing et al. 2020). The highly modular nature of AI-based knowledge graphs, an example of which has been shown in figure 3, posit these as versatile tools for efforts aimed at rapid drug prediction from extensive libraries. A similar report was published which used a drug repositioning framework employing ML and statistical analysis. The results were validated by testing its predictions for MERS-COV and SARS-COV, which were shown to be in line with the experimentally observed results (Ge et al. 2020). A reverse vaccinology study employing Vaxign-ML (a machine learning-based application) was recently carried out (Ong et al. 2020). The basic workflow involves a bioinformatic analysis of the genome/proteome, based upon which suitable targets are scored. Given the number of existing and reported studies targeting the viral spike protein (which also received the highest score using Vaxign-ML), the authors of this study decided to focus on the nsp3 which was ascribed the second-highest score by the aforementioned algorithm. It appears that vaccine candidate identification efforts stand to benefit from such ML-driven analysis, which may have an inherent edge in unearthing unconventional and generally non-obvious targets offering similar or superior outcomes as compared to those determined using traditional workflows or established tenets.

Figure 3
figure 3

An example illustrating the highly modular nature of AI-based knowledge graphs.

Full size image

5 Vaccine development for SARS-CoV-2

Considering the highly contagious nature of the SARS-CoV-2 infection, the only long-term sustainable solution is the development of robust vaccination protocols. In this brief section, we wish to draw the reader’s attention to some innovative, cutting-edge approaches towards determining potential vaccine candidates.

An approach based on comparing the human and viral proteomes to search for pentapeptide sequences unique to the SARS-CoV-2 was proposed by Guglielmo Lucchese, as these would be expected to have high immunogenicity as well as specificity (Lucchese 2020). Immunoinformatics-guided investigations may also play a significant role in the identification of suitable target epitopes, as shown by two recent papers (Ahmed et al. 2020; Baruah and Bose 2020). Baruah and Bose screened the SARS-CoV-2’s surface glycoprotein for CTL and B-cell epitopes, further supplemented by molecular dynamics simulations to adjudge the interactions of the former with MHC Class I types majorly represented in the Chinese population (Baruah and Bose 2020). The work by Ahmed et al. incorporated insights from the study of immunogenic SARS-CoV proteins and B/T-cell epitopes derived from the same to obtain a set of such epitopes for the SARS-CoV-2, which have also been verified to be invariable amongst the published SARS-CoV-2 genomes, thereby making these attractive targets (Ahmed et al. 2020). Another notable effort undertook an in-silico approach aimed at designing a multi-epitope vaccine candidate by predicting B/T-cell epitopes based on analysis of the viral nucleocapsid, ORF3a, and membrane protein (Enayatkhani et al. 2020). It is conceivable that a comprehensive screen of epitopes identified in the above-cited (and other related) literature in suitable animal models may yield lead candidates for vaccine trials.

Data generated by previous studies on the SARS-CoV significantly facilitates current vaccine development efforts against the SARS-CoV-2. For instance, an analysis of the antigenic and glycosylation variation between SARS CoV and SARS-CoV-2 revealed that while several novel glycosylation sites were observed in the receptor-binding domain of SARS-CoV-2, the structure of the glycoprotein showed no significant divergence (Kumar et al. 2020). Additionally, despite the antigenic variability between both the strains, certain CTL epitopes were found to be conserved, suggesting that existing peptide-based vaccine candidates against the SARS-CoV could be repurposed for testing against SARS-CoV-2. This observation is supplemented by a recently published report, wherein Conformational-Epitope BLAST (which identifies antigenic similarity of a new pathogen to existing ones) was used to identify cross-reactive epitopes in the ACE2 receptor binding region of SARS-CoV-2’s S protein (Qiu et al. 2020). Along similar lines, polyclonal antibodies against the SARS-CoV receptor-binding domain have been shown to cross-react with SARS-CoV-2 receptor-binding domain (RBD), thereby inhibiting SARS-CoV-2’s entry into cells (Tai et al. 2020). A largely concordant line of reasoning has also advocated the consideration of the RBD219-N1 (a yeast-expressed potential vaccine candidate against SARS-CoV) as a heterologous vaccine against the SARS-CoV-2 (Chen et al. 2020b). These examples serve to highlight that alongside initiatives to develop novel vaccine candidates, repurposing efforts could help provide immediate solutions to address the urgent needs imposed by the ongoing pandemic situation.

A cell-based antigen delivery system against the SARS-CoV-2 has also been proposed recently using decoy cells displaying viral spike protein (Ji et al. 2020). The report provides evidence from the literature to support and justify the approach, in particular, citing the example of GM-CSF expressing irradiated cells being well tolerated in clinical trials against cancer. However, no records of assays or in-vivo animal model studies have been provided at the moment; hence it is premature to comment on the suitability of this concept in the context of SARS-CoV-2 infections.

The recent review by Iwasaki and Yang raises important caveats regarding vaccine development efforts, in particular, the phenomenon of antibody-dependant enhancement (ADE). ADE occurs when therapeutically administered antibodies can exacerbate the existing pathology by enabling infection of immune cells by the invading virus, promoting inflammation, downregulation of anti-inflammatory factors like IL-10, and aberrant activation of the immune system leading up to tissue injury. The dangers associated with hyper-inflammation and immune dysregulation in the context of SARS-CoV-2 infections have been elucidated at length in Sect. 3.4. The authors outline several possible contributing factors towards ADE, including antibody specificity, choice of target epitope, affinity, subtype (IgM or IgG), dosage, and patient status. Such concerns underscore the need for well-rounded vaccine development protocols and extensive preclinical evaluations in order to optimize the choice of adjuvants, mode of administration, and other relevant parameters as enlisted above. A more thorough elaboration may be found in the publication referenced (Iwasaki and Yang 2020).

The World Health Organization (WHO) is currently maintaining publicly accessible, comprehensive, and authoritative databases (https://www.who.int/publications/m/item/draft-landscape-of-covid-19-candidate-vaccines) listing all the vaccines under clinical or preclinical evaluation (WHO 2020a). At the time of writing, 33 vaccine candidates are under clinical trial, which have been summarized in table 2. Additionally, the WHO’s COVID-19 modeling ad hoc expert working group (https://www.who.int/publications/m/item/covid-19-animal-models---summary-of-progress-made-by-the-who-covid-19-modelling-(march-04-june-2020)) has also reported (based on the available results from studies across the world) that Rhesus macaques and ferrets appear to be the most suitable large animal models for studying SARS-CoV-2 infections to date (WHO 2020). Such public information initiatives are of immense benefit to encourage collaboration, avoid unnecessary and wasteful duplication of effort, and disseminate valuable data rapidly across the research community. We would also like to direct the interested reader to refer to an excellent summary of the SARS-CoV-2 vaccine landscape by Nature Reviews Drug Discovery and authors affiliated with the Coalition for Epidemic Preparedness Innovations (CEPI) (Le et al. 2020).

Table 2 Vaccines being developed against COVID-19 currently in clinical trial stages
Full size table

6 Conclusion

As the current progress of the SARS-CoV-2 outbreak has made obvious, clinical management of these infections has been under multiple pressures ranging from high false-negative rates of diagnosis, contentious results on the efficacy of proposed drugs, and the threat of severely detrimental symptoms like ARDS and cytokine storm responses. The development of accessible point-of-care diagnostic devices tailor-made for low-resource settings continues to be a pressing requirement. This may be achieved by cell lysis-free workflows, increasingly sensitive and robust bioassays, improved sample collection and processing methods which minimize risks to healthcare workers, and ready-to-use kits not requiring technical expertise, among other possible innovations. In this regard, RT-LAMP and CRISPR/Cas-based methods like SHERLOCK and DETECTR are notable advancements that could be explored as promising alternatives. With regard to portable diagnostic devices, few technologies hold as much promise as paper-based microfluidics. The applicability of this has already been demonstrated in the context of the Ebola and Zika outbreaks (Kaarj et al. 2018; Magro et al. 2017). While a comprehensive overview of the microfluidics domain is beyond the scope of this work, considering the immense potential of this technology we would like to direct the interested reader to refer to a dedicated review of paper-based microfluidic platforms for nucleic acid detection (Kaur and Toley 2018).

Advocating accelerated vaccine development efforts might prove to be a risky gamble, on account of the associated safety caveats which have been underscored in the preceding sections. Instead, an emphasis on designing and adopting reliable biophysical assays for comprehensive, yet rapid preclinical evaluation may be beneficial in reducing the historically high attrition rate among candidates under clinical trials. This holds especially true for studies proposing novel therapeutic agents as well, for which emerging techniques like the CETSA (cellular thermal shift assay) may help deter the occurrence of misleading false positives (Martinez Molina et al. 2013). Furthermore, as regards the application of AI/ML-based platforms in biomedical research and diagnostics, a confluence of cross-domain expertise is crucial to ensure the development of reliable models and workflows which faithfully capture real-world dynamics. As instances of governments and healthcare agencies employing AI-based containment and prevention strategies increase, it is important that critical studies be undertaken to ascertain the true benefits and outcomes of such technological integration, rather than a reliance on anecdotal evidence alone.

Throughout our discussion, certain points have been recurrent – the advantages of leveraging existing knowledge on closely related species like the SARS CoV, the potent efficacy of seemingly disparate techniques when applied in tandem, and the importance of collaborative efforts coupled with rapid, timely dissemination of results which may inform ongoing efforts. Requirements key to enabling these include the availability of genomic sequences, structural data, and expertly curated databases. The push towards developing modular assay workflows (with elements of automation involved) may well be a defining paradigm of future responses to such scenarios. The advent of in silico epitope screening, algorithmic vaccine and inhibitor design, and preliminary evaluation of drug libraries by molecular docking has the potential to greatly facilitate wet-lab-based and clinical investigations while expediting results. However, we reiterate our concerns over adequate quality testing and rigorous experimental validation before adopting such approaches as an established norm. Skepticism must also be extended in light of a growing tendency to extrapolate trends observed from in vitro setups and unsuitable animal models to recommend clinical guidelines. We conclude our narrative on the hopeful note that the topical coverage should have appealed to both experts in virology and biomedical engineers, as well as recent entrants in these research domains. The purpose of this effort would be best served if the diverse studies described herein motivate and help lay the foundation for ambitious multidisciplinary undertakings aimed at the advancement of public health.