Standardizing antiviral response metrics for mono- and combination therapies in acute, chronic and latent viral infections

Abstract

The COVID-19 pandemic showed that heterogeneous antiviral assay designs, endpoints and reporting practices can obscure which candidate drugs and combinations are genuinely promising. As antiviral discovery expands from acute infections to chronic and latent viral diseases, the field needs a compact, reproducible and biologically interpretable set of response metrics. In this Personal View, we argue that EC₅₀ and the selectivity index should be retained for pharmacological interpretation but systematically complemented by drug sensitivity scores (DSS), which integrate potency and efficacy across the tested concentration range, and by ΔDSS, calculated as DSS_antiviral minus DSS_toxicity from matched efficacy and viability curves. We propose standardized modules for resistance passaging, sequencing and host-side-effect profiling so that monotherapies are assessed not only for antiviral potency but also for durability and host-cell perturbation. For combinations, we discuss Bliss, ZIP, HSA and Loewe models as complementary tools for identifying additive or synergistic regimens while accounting for toxicity, resistance suppression and drug-drug interactions. Finally, we outline how harmonized metrics can support organoid, animal and clinical prioritization, improve machine-learning datasets and guide pandemic response, including rapid monotherapy testing for some DNA viruses and early combination testing for RNA and reverse-transcribing viruses.

Introduction

High-throughput screening and drug repurposing have become central strategies for rapid antiviral discovery [1, 2]. During the SARS-CoV-2 pandemic, hundreds of candidate monotherapies and combinations were tested in cell lines, primary cells and human organoids [3, 4]. However, the resulting data were often difficult to compare because studies used different viral strains, host systems, exposure times, readouts and quantitative metrics [5, 6]. Some reports relied on single-concentration inhibition values, others reported EC₅₀ values without matched toxicity measurements, and relatively few applied formal synergy or resistance metrics [7, 8]. This heterogeneity can overstate the promise of compounds that later fail in vivo or clinically, and it limits the reuse of preclinical data for meta-analysis and machine learning.

Curated resources and analysis platforms show that this problem is solvable. DrugVirus.info integrates data on broad-spectrum antiviral agents (BSAs) and BSA-containing combinations across virus families and host models when assays are annotated and analysed through a common pipeline [9]. Breeze and SynergyFinder provide practical tools for calculating DSS and synergy scores from dose-response data [8, 10]. Together, these resources provide a basis for harmonizing antiviral response metrics across acute, chronic and latent infections, while also making the assumptions behind each score transparent.

We organize the proposed framework into four linked steps. First, a single antiviral candidate should be evaluated using standardized monotherapy metrics, including EC₅₀, CC₅₀, SI, DSS_antiviral, DSS_toxicity and ΔDSS. Second, promising monotherapies should be tested for resistance emergence and host-side effects using serial passaging, sequencing, viability and host-response assays. Third, if monotherapy activity is incomplete, resistance emerges or broader coverage is desired, the candidate can be tested in combination with existing BSAs or other antiviral candidates using standardized synergy models. Fourth, in a pandemic-preparedness setting, the balance between monotherapy and combination therapy should be guided by viral biology: monotherapy may be a practical first option for some DNA viruses, whereas RNA viruses and reverse-transcribing viruses often require early combination strategies because rapid mutation and genetic diversity can favour resistance to single agents.

Monotherapy response metrics

EC₅₀ remains the most widely reported antiviral potency metric [11]. It is usually derived from a sigmoidal concentration-response model, often a four-parameter log-logistic fit, using a virological endpoint such as viral RNA, infectious titre or reporter signal [12]. A matched CC₅₀ is obtained from a parallel viability or cytotoxicity assay, and SI = CC₅₀/EC₅₀ is used as an approximate in vitro therapeutic-window metric [13]. These parameters are familiar and clinically interpretable, but they summarize only selected features of the response curve.

EC₅₀ and SI have important limitations. EC₅₀ can shift when maximal inhibition is incomplete, slopes are shallow, assay noise is high or the tested concentration range does not cover the full response [14]. SI compounds uncertainty because it combines two fitted quantities; in particular, CC₅₀ is often extrapolated beyond the highest tested concentration, which can inflate perceived safety. These instabilities are problematic for cross-study ranking and for machine-learning models that require robust numerical features [15].

DSS addresses some of these limitations by summarizing the area of biologically meaningful activity across the dose-response curve rather than relying on a single concentration [16]. By integrating both potency and efficacy above a predefined activity threshold, DSS can distinguish a compound with broad, consistent antiviral activity from one that shows inhibition only at the highest concentration, where off-target toxicity may also appear [17].

For reproducibility, DSS calculation should be described algorithmically. After fitting or interpolating the dose-response curve, the activity above a predefined threshold t is integrated over the tested log-concentration range and normalized to the maximum possible area. A practical formulation is:

where t is the activity threshold, cmin and cmax define the tested concentration range, and AUC_threshold represents the baseline area below the threshold. DSS is commonly scaled from 0 to 50, with higher values indicating stronger and broader activity. Because DSS integrates the active portion of the full dose-response curve, it is often more robust than EC₅₀ when curves differ in slope or do not reach a clear plateau [18].

DSS is especially useful for broad-spectrum antiviral discovery because the same calculation can be applied across viruses, cell types and readouts when the concentration range, threshold, fitting model and curve quality criteria are reported [9, 19]. Nevertheless, DSS is not independent of experimental design. It should therefore be interpreted together with EC₅₀, maximal effect, curve slope, assay timing and the clinically achievable exposure range rather than as a stand-alone efficacy metric.

For antiviral development, DSS should be calculated separately for antiviral activity and toxicity using matched dose-response curves generated under comparable experimental conditions. This enables calculation of:

ΔDSS places antiviral activity and toxicity on the same numerical scale and provides a compact measure of net in vitro selectivity. High positive ΔDSS values indicate strong antiviral activity with limited toxicity across the tested range; values near zero suggest a narrow separation between efficacy and toxicity; and negative values indicate that toxicity outweighs antiviral benefit. Figure 1A illustrates how EC₅₀, CC₅₀, SI, DSS_antiviral, DSS_toxicity and ΔDSS can be reported together for a single drug candidate.

Fig. 1.

Standardized workflow for preclinical development of Antiviral X. (A) Antiviral X dose-response analysis. Antiviral activity and matched cell viability are measured across a concentration range to calculate EC₅₀, CC₅₀, selectivity index (SI), DSS_antiviral, DSS_toxicity and ΔDSS. These metrics provide an integrated assessment of potency, efficacy, toxicity and in vitro selectivity. (B) Organoid and ex vivo testing. Lead concentrations or dosing regimens are advanced to organoid and ex vivo tissue-slice models derived from virus-susceptible tissues. Representative high-content imaging can be used to assess infection, antiviral activity, tissue-level response and cytotoxicity. (C) In vivo testing. Antiviral X is evaluated in animal models using defined treatment schedules relative to virus infection. Endpoints may include survival, body weight, viral load, tissue pathology, inflammatory cytokines and safety/toxicity. (D) Development timeline and approximate costs. After early discovery and lead identification, Antiviral X progresses through preclinical evaluation, IND-enabling studies and clinical trials. The indicated times and costs are approximate and may vary substantially depending on the viral target, model system, regulatory requirements, formulation, manufacturing complexity and whether the compound is a new chemical entity or a repurposed drug

The therapeutic window remains a critical complementary concept. A compound with a favourable in vitro DSS profile may still have limited translational value if antiviral activity occurs only at concentrations that cannot be reached safely in plasma or in the relevant tissue. Therefore, EC₅₀, CC₅₀, SI, DSS and ΔDSS should be considered alongside pharmacokinetics, pharmacodynamics, protein binding, tissue penetration and tolerated dosing regimens.

Minimum reporting standards for monotherapy assays should include the virus strain and multiplicity of infection, host-cell or organoid model, assay duration, endpoint, concentration range, number and spacing of doses, activity threshold used for DSS, fitting or interpolation method, replicate number, curve-quality criteria and whether toxicity was measured on the same plate or in a matched parallel format. These details are essential for reproducible DSS and ΔDSS implementation.

We therefore propose that EC₅₀ and SI should not be abandoned, but should be complemented routinely by DSS_antiviral, DSS_toxicity and ΔDSS. This preserves familiar pharmacological interpretation while providing a more robust framework for ranking monotherapies across viruses, cell types and assay platforms.

Acute infections: short-term antiviral responses and harmonisation

Acute viral infections such as influenza, respiratory syncytial virus infection, coronavirus infection and enterovirus infection are characterized by rapid replication and host responses over hours to days [20, 21]. In vitro assays typically measure single- or limited multi-round replication 24–96 h after infection [22–24]. Under these conditions, antiviral activity and matched viability can usually be measured in the same experiment or in a closely matched parallel format, allowing EC₅₀, CC₅₀, SI, DSS_antiviral, DSS_toxicity and ΔDSS to be computed within one analytical pipeline [25].

For acute infections, we recommend reporting all six monotherapy quantities but using DSS and ΔDSS as primary ranking variables for cross-study comparison. This is particularly useful when transformed cell lines, primary cells and organoids are tested in parallel, because absolute EC₅₀ values may shift owing to differences in uptake, metabolism or basal antiviral state, whereas the area-based framework can still support standardized ranking [25–27].

Latent and chronic infections: additional dimensions of response

Latent and chronic viral infections, including herpesvirus, hepatitis B virus, HIV and polyomavirus infections, present a broader response space than acute infections [28–30]. Short-term suppression of active replication is only one therapeutic goal; durable suppression, reservoir reduction, prevention of reactivation and preservation of host-cell function may be equally important [31, 32]. We therefore treat the application of DSS to latency and reservoir assays as a forward-looking extension rather than as an established universal standard [33].

A practical way forward is to define endpoint-specific DSS values. In hepatitis B virus models, DSS_suppression could be calculated from extracellular HBV DNA, HBsAg or HBeAg decline during repeated dosing, whereas DSS_reservoir could be calculated from cccDNA-associated or other reservoir-relevant endpoints after treatment or washout [34]. In HIV latency models, DSS_reactivation could be calculated from GFP-positive, p24-positive or transcriptionally reactivated cells after a defined stimulus, with DSS_toxicity measured in matched controls [35]. The key requirement is that each endpoint be explicitly named, biologically justified and analysed with the same mathematical framework so that latent-virus assays generate structured, machine-readable data while acknowledging that consensus endpoints are still evolving [36, 37].

Standardization of antiviral resistance and side-effect testing

Two major challenges accompany antiviral development: resistance emergence and host-side effects [38–42]. Resistance arises when replication under drug pressure selects variants with reduced susceptibility. This problem is documented across antiviral classes; for example, SARS-CoV-2 evolution has compromised the activity of several previously useful neutralizing antibodies and small-molecule inhibitors [38]. More generally, resistance can emerge through target-site mutations, altered viral entry pathways, compensatory fitness changes or incomplete suppression during prolonged treatment [38, 42].

We therefore propose that antiviral evaluation should include a standardized resistance module alongside EC₅₀, CC₅₀, SI, DSS and ΔDSS (Fig. 2A). Parental virus should be serially passaged under defined Antiviral X pressure and in parallel without drug pressure to distinguish drug-selected escape from background culture adaptation. Viral replication should be monitored at each passage, and standard endpoints should include time to viral rebound, number of passages to escape, fold-change in susceptibility measured by EC₅₀ and/or DSS, resistance-associated mutations and the fitness of emerging variants in the absence and presence of drug. This approach builds on work showing that antiviral combinations can delay resistance and should be incorporated into preparedness-oriented pipelines [42, 43].

Fig. 2.

Standardized assessment of antiviral resistance and host-side effects during antiviral development. (A) Resistance passaging and sequencing. Parental virus is subjected to serial passaging under and without antiviral drug pressure to monitor the emergence of escape variants and to distinguish drug-selected mutations from background adaptation during culture. Viral replication is followed across passages, and viral populations are analysed by deep sequencing to identify escape mutations and their frequencies. Key outputs include time to viral rebound, fold-change in drug susceptibility measured by EC₅₀ and/or DSS, resistance-associated mutations and the fitness of emerging variants in the absence and presence of drug pressure. (B) Side effects and host-response profiling. Antiviral-treated mock-infected and infected cells or organoids are compared to distinguish antiviral activity from host-cell perturbation. Primary assays assess cell viability, cytotoxicity, apoptosis, stress markers, barrier function and cytokine or chemokine release. These readouts can be complemented by transcriptomic, proteomic, phosphoproteomic and metabolomic profiling, with AI-assisted analyses used to predict affected pathways, toxicity mechanisms and drug-drug interaction risks

A parallel side-effect module is also needed (Fig. 2B). Cell viability alone is insufficient to exclude biologically important adverse effects, especially for host-targeted antivirals, immunomodulators and combinations [39–41]. ABT-263 accelerated apoptosis in virus-infected cells, altered cytokine production and reduced survival in infected mice [40]. JAK inhibitors suppressed innate immune barriers and facilitated viral propagation, illustrating that anti-inflammatory activity can have unintended pro-viral consequences [39]. Commonly prescribed medicines can also alter influenza A virus-host interactions at transcriptomic and metabolic levels [41].

Side-effect testing should therefore compare mock-infected and infected cells or organoids treated with the same monotherapy or combination. Primary readouts should include viability, cytotoxicity, apoptosis, stress markers, barrier integrity where relevant and cytokine or chemokine release. These assays can be complemented by targeted host-response profiling and, when justified, transcriptomics, proteomics, phosphoproteomics and metabolomics. AI-assisted analyses may help prioritize toxicity pathways, drug-drug interactions and vulnerable host networks, but such predictions require experimental validation. Together, standardized resistance and side-effect modules identify regimens that are potent, durable and less likely to produce harmful host perturbations [38–42].

Paths to standardization in organoid-based antiviral testing

Standardization is essential if organoid and ex vivo data are to contribute quantitative evidence rather than isolated validation images (Fig. 1B). Minimum information should include organoid source and differentiation protocol, maturation state, infection conditions, endpoint timing, viability threshold, imaging or molecular readout and whether concentration-response designs were used. Automated imaging and AI-assisted quantification can support DSS calculation when multiple concentrations are feasible. When throughput is limited, organoids should still be used to confirm whether the concentration range identified in simpler models produces antiviral activity without unacceptable tissue-level toxicity [3, 4, 24–27].

Harmonised in vivo testing of antiviral therapies

Standardized in vitro metrics are necessary but not sufficient for rational antiviral development (Fig. 1C). Before clinical testing, promising monotherapies should be evaluated in vivo using harmonized protocols that allow comparison across drugs, viruses and laboratories [44]. Key design variables include viral strain and inoculum, route of infection, animal age and sex, timing and route of drug administration, humane endpoints, sampling schedule and pharmacokinetic measurements.

For acute infections, primary readouts should include survival, body weight or clinical score, viral load in relevant tissues or secretions, pathology, inflammatory cytokines and safety/toxicity. Virological endpoints should be measured with harmonized assays, including clear limits of detection and standard curves. Where feasible, pharmacokinetic sampling should confirm that in vivo exposures overlap the concentrations at which ΔDSS was observed in vitro.

Latent and chronic infections require additional in vivo endpoints, including reservoir size, frequency and severity of reactivation, longitudinal safety and functional immune readouts. Because such studies are longer and more resource-intensive, standardized protocols and common data elements are especially important for aggregating results across laboratories and for modelling translational potential.

Clinical trials and translational implications: cost, complexity and prioritisation

Preclinical standardization is useful only if it connects coherently to clinical development (Fig. 1D) [45]. The transition from in vitro ΔDSS, resistance data, side-effect profiles and animal efficacy to human benefit is constrained by biology, regulatory requirements, manufacturing, cost and trial feasibility. Therefore, standardized metrics should be used as triage tools: they should identify which candidates justify the investment required for IND-enabling studies and clinical trials.

The most promising monotherapies should combine robust antiviral DSS across relevant host systems, a positive ΔDSS, delayed resistance emergence, low host-side-effect burden and activity at clinically achievable concentrations.

Combination therapy: synergy scores

Combination therapy is central to antiviral treatment because it can increase efficacy, reduce the required dose of each component and delay resistance, but no single interaction model is appropriate for every biological setting [19]. Several reference models are widely used, including Highest Single Agent (HSA), Loewe additivity, Bliss independence and Zero Interaction Potency (ZIP) [46]. HSA is useful when the question is whether the combination outperforms better monotherapy at the same concentrations. Loewe is informative for drugs with similar or partially overlapping mechanisms, although it can be unstable when monotherapy curves are shallow or poorly fitted. Bliss and ZIP are practical defaults for antiviral matrix screens and are implemented in widely used tools [8, 47].

As shown in Fig. 3, combination development should be separated from the initial single-drug workflow. A candidate such as Antiviral X can be screened with existing BSAs or other antiviral candidates in dose-response matrices to identify additive or synergistic activity using ZIP, Bliss, HSA or Loewe models [8, 46, 48–51]. The same screen should evaluate cytotoxicity, host-side effects, drug-drug interaction risk and clinically achievable exposure ranges. This is particularly important when monotherapy activity is incomplete, dose escalation is limited by toxicity or resistance emerges during serial passaging [42, 43].

Fig. 3.

Framework for antiviral drug-combination development. (A) Dose-response matrix and ZIP synergy analysis. Antiviral combinations are first evaluated in a concentration-response matrix to quantify antiviral activity across dose pairs. In the example shown, remdesivir and alvespimycin hydrochloride are tested in combination, and antiviral activity is summarized as a response heatmap. The corresponding ZIP synergy map identifies regions of positive or negative interaction across the tested concentration range and helps define dose pairs that produce additive or synergistic antiviral effects. (B) Strategy for antiviral drug-combination development. A newly available antiviral, represented here as Antiviral X, is first evaluated in cell culture for antiviral efficacy, toxicity and resistance emergence. If resistance emerges, or if improved activity is needed, Antiviral X can be screened together with existing BSAs to identify combinations with favourable antiviral activity, including additive or synergistic effects, while also considering side effects and potential drug-drug interactions. Promising combinations can then be advanced to organoid models, animal models and ultimately clinical trials. The timeline and cost estimates shown are approximate and illustrate the relative scale of progression from early screening to clinical development

Bliss independence assumes that two drugs act independently on the same population of infected cells or targets [52]. The expected fractional remaining infection or viability is the product of the individual remaining fractions; synergy is observed when the combination effect exceeds this expectation, and antagonism when it falls below it. In antiviral studies, Bliss is intuitive when mechanisms are distinct, such as combining a polymerase inhibitor with a host-targeted entry blocker [53].

ZIP incorporates elements of Bliss and Loewe by asking whether one drug changes the apparent potency of the other across the response surface [48]. It is therefore well suited for matrix-wide comparison, because it detects systematic shifts in potency or efficacy rather than only pointwise deviations. ZIP scores can be summarized for the full matrix or for clinically relevant concentration windows, making them useful for ranking large numbers of combinations.

SynergyFinder and related tools calculate Bliss, ZIP and other synergy scores for dose combinations and summarize them across matrices [8, 49]. DrugVirus.info applies such pipelines to antiviral combination datasets, enabling comparison of broad-spectrum antiviral combinations across viruses and host models [9]. Integrated approaches such as SynToxProfiler extend this logic by considering synergy, efficacy and toxicity together, which is essential when infected and matched control-cell matrices are available [50, 51].

Higher-order combinations can be analysed by extending the same principles to multidimensional dose-response landscapes, but interpretability and experimental cost increase rapidly. Such combinations should be prioritized only when there is a strong mechanistic rationale, when pairwise data support consistent selectivity and when toxicity and drug-drug interaction risks remain manageable [43, 50, 51].

In routine analysis, negative Bliss or ZIP scores indicate antagonism, positive scores indicate synergy and values close to zero indicate additivity or no detectable interaction. Common working thresholds are approximately δ < -10 for antagonism, -10 <= δ <= 10 for additivity or inconclusive interaction and δ > 10 for meaningful synergy. These thresholds are heuristic rather than universal; confidence intervals, replicate consistency, assay noise, toxicity and clinical exposure windows should always be considered (Fig. 3).

For acute infections, Bliss and ZIP can serve as a practical default pair: Bliss provides a simple mechanistic expectation at defined concentration pairs, whereas ZIP provides a surface-level summary that is useful for ranking and machine learning. However, all synergy estimates can be distorted by strong cytotoxicity, incomplete dose-response coverage or poorly fitted monotherapy curves. Combination studies should therefore report the full matrix, monotherapy curves, selected reference model, toxicity in matched control cells and the concentration window used for summary scores.

For latent and chronic infections, combination models can be applied to endpoint-specific matrices such as long-term suppression, reactivation frequency or reservoir reduction. Synergy in short-term replication suppression may not correspond to synergy in reservoir reduction, so endpoints should be analysed separately before being integrated into a development decision.

Translating in vitro synergy into animal studies is not straightforward because in vivo dose matrices are necessarily sparse. Nonetheless, structured designs can include at least two dose levels for each monotherapy and the corresponding combination doses, analysed with factorial or response-surface models when feasible. The primary question should be whether the combination improves survival, clinical score, viral-load reduction, resistance suppression or safety at clinically realistic exposures, rather than whether it reproduces an exact in vitro synergy score.

Combination development also increases cost and complexity faster than monotherapy development [43]. Each additional component increases the number of possible doses, schedules, sequences and interaction liabilities. Standardized preclinical metrics are therefore not merely descriptive; they are triage tools that help prevent weak or poorly selective combinations from entering expensive animal studies or clinical trials.

Clinical trials can partly address this complexity through randomized, stratified, adaptive or platform designs [45]. For acute infections, such designs can drop ineffective arms and prioritize regimens with early virological benefit. For chronic and latent infections, master protocols may evaluate sequences or cycling of regimens informed by preclinical metrics on suppression, reactivation and toxicity. However, efficient trial designs cannot compensate for indiscriminate upstream candidate selection.

Outbreaks caused by emerging or re-emerging viruses add logistical and ethical constraints. Sporadic outbreaks may involve small patient numbers, remote settings and limited capacity for standardized sampling. For chronic infections, existing regimens and intellectual-property constraints may limit how easily new combinations can be tested or deployed. These realities reinforce the need for transparent, harmonized preclinical prioritization before clinical development.

The practical conclusion is that combinations should be advanced only when they offer a clear benefit over monotherapy: stronger or broader antiviral activity, lower required doses, delayed resistance, acceptable toxicity, manageable drug-drug interactions and realistic clinical exposure.

Harmonised metrics and machine learning for pandemic preparedness

Machine-learning methods are increasingly used to predict antiviral candidates, prioritize drug combinations and extrapolate activity to new viruses from chemical structure, host response and prior pharmacology [27]. These methods require large, consistently annotated datasets. Heterogeneity in assay design, endpoints and metrics currently limits model generalization across viruses, cell types and laboratories. From a modelling perspective, a smaller set of well-defined features, such as EC₅₀, DSS, ΔDSS and one or two standardized synergy scores, is more useful than a larger set of idiosyncratic readouts.

Resources such as DrugVirus.info show how standardized annotation of EC₅₀, CC₅₀, SI, DSS and synergy scores can create cleaner training data by linking drug responses to virus families, host-cell models and clinical development status [9]. Similar principles should be applied to organoid, ex vivo, animal and resistance datasets. Harmonized metrics will not by themselves guarantee accurate prediction, but they reduce avoidable technical heterogeneity and improve the likelihood that models generalize to unseen viruses and host systems [9, 10, 27].

From a pandemic-preparedness perspective, harmonization should translate into a standardized early-response pathway (Fig. 4). After a new virus outbreak, virus identification should be followed by rapid testing of existing or repurposed antiviral candidates. For some DNA viruses, a repurposed Antiviral X may provide a practical early monotherapy option. For RNA viruses and reverse-transcribing viruses, Antiviral X should be tested rapidly in BSA-containing combinations because rapid replication, error-prone polymerases, reverse transcription and high genetic diversity can promote resistance to monotherapy [20, 28, 29, 42, 43]. In parallel, neutralizing antibodies and vaccines can be developed over longer timelines. Standardized monotherapy and combination testing therefore provides an early pharmacological bridge while virus-specific antibodies and vaccines are being developed.

Fig. 4.

Standardized pandemic response to a new virus outbreak. After a new virus outbreak, virus identification is followed by rapid testing of repurposed Antiviral X for DNA viruses, or Antiviral X in combination with existing BSAs for RNA and reverse-transcribing viruses. In parallel, neutralizing antibodies and vaccines are developed over longer timelines. Drug combinations are particularly important for RNA and reverse-transcribing viruses because their high mutation rates and genetic diversity can promote rapid resistance to monotherapy. Standardized combination testing can therefore provide an early pharmacological bridge while virus-specific antibodies and vaccines are being developed

Conclusions

The next pandemic will again require rapid identification of antiviral monotherapies and combinations, not only for acute respiratory pathogens but also for chronic and latent viruses that may reactivate during stress, immunosuppression or co-infection. Preparedness will be improved if the field moves beyond ad hoc reporting and adopts a compact, reproducible set of response metrics connected to resistance, side-effect and translational testing. For monotherapies, EC₅₀ and SI should be retained for interpretability but systematically complemented by DSS_antiviral, DSS_toxicity and ΔDSS, calculated with clearly reported concentration ranges, thresholds and curve-quality criteria. For combinations, Bliss and ZIP provide practical default synergy models, while HSA and Loewe remain important comparators for specific mechanistic questions. Combination ranking should also include toxicity, host-side effects, resistance suppression, drug-drug interactions and clinically achievable exposure. Databases and tools such as DrugVirus.info, Breeze, SynergyFinder and SynToxProfiler already provide much of the necessary infrastructure [8–10, 49–51]. By agreeing now on how to quantify, report and interpret antiviral effects in acute, chronic and latent infections, the community can build datasets that are directly usable for rational prioritization, machine learning and faster pandemic response.

Author contributions

Conceptualization, D.K.; writing—original draft preparation, D.K.; writing—review and editing, all authors.; visualization, O.K., D.K., E.R., V.D.; supervision, D.K.; project administration, D.K. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no specific grant from any funding agency in the public, private, or not-for-profit sectors.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.