What marine virus-like particle counts actually describe
A seawater report may state that one millilitre contains ten million virus-like particles. That figure is not a direct count of infectious viruses, identified viral species or cells that died. Epifluorescence microscopy counts fluorescent points meeting an operational definition after a defined sample volume is filtered, stained and examined. Some points may be noninfectious particles, while some particles may be missed or merged with background. The strength of this method is its ability to estimate abundant tiny particles efficiently; its limitation is that a fluorescent dot is not a complete biological identity.
Marine viruses play roles in microbial ecology and element cycling, so abundance measurements are important. Yet abundance is a standing inventory, not a direct production rate, infection frequency or lysis rate. A large count can coexist with rapid loss or slow turnover, just as a crowded station can have very different passenger flows. To interpret a result, begin with sample handling, filter properties, stain, microscope settings, effective counted area and statistical treatment rather than moving directly from count to ecosystem impact.
This article develops a measurement-first reading method. It explains what SYBR-stained particles reveal, how a filter field becomes a concentration, how blanks and recovery affect the result and what the data cannot prove. The arithmetic examples are explicitly hypothetical teaching exercises. They are designed to help readers reconstruct published calculations without pretending that a fictional slide or ocean station was actually observed.
Why optical microscopy can reveal particles smaller than its resolving power
Many marine viruses are too small for an ordinary light microscope to resolve as detailed structures. Fluorescent nucleic-acid stains can nevertheless make sub-resolution particles appear as bright points of light through their emitted signal. The microscope does not thereby reveal the viral capsid’s shape or genome. It detects a fluorescent event, sometimes smaller than the optical point-spread function. Counting such points is possible when the signal, focus and background are sufficient to distinguish individual events.
Established methods have used dyes such as SYBR Green I and SYBR Gold and collected particles on very fine-pore filters, including 0.02-micrometre aluminium oxide membranes in published protocols. The exact preparation affects the signal and background. Different dyes, mounting solutions, optics and image processing may yield different counts from the same sample. One method’s fluorescence threshold cannot be assumed interchangeable with another method’s threshold without a comparison study.
The descriptive name ‘virus-like particle’, or VLP, deliberately preserves the operational limit. A point that stains with a nucleic-acid dye and resembles a viral particle under the counting criterion is not automatically a cultured infectious virus. A total count may also include background events depending on the preparation and analysis. Readers should treat VLP abundance as a validated proxy for particle abundance, not as species identification or a pathogen diagnostic test.
Before counting: preserve the sample and its provenance
Seawater contains living cells, free particles, detritus, colloids and dissolved organic matter. Sampling and handling can alter how those components distribute. The record should preserve station, collection time, water depth or pressure, container, treatment, storage interval, dilution, filtration volume and staining method. A short label such as ‘surface sample’ is insufficient if later analysts want to compare counts from different cruises or identify possible handling effects.
Some protocols use preservation and freezing steps, but the effectiveness depends on the validated method. Uncontrolled delay can change particles through biological processes, adsorption, aggregation or degradation. Rough pipetting or filtration may introduce artifacts, and highly turbid water can overload the filter. When comparing preserved and fresh samples, the method must establish that particle recovery and staining remain comparable. A successful final microscopic image cannot tell the reader what happened before filtration.
Sampling volume matters because the count is eventually divided by that volume. Record the actual volume passing through the membrane, not merely the nominal amount placed in a tube. If a filter clogs or liquid remains in the apparatus, the processed volume may differ. For diluted samples, keep both the measured aliquot volume and the original-water dilution factor. Many large apparent changes in environmental abundance can arise from a simple missing factor in this part of the record.
Filtration concentrates particles, but it also selects them
A fine-pore filter collects particles from a measured water volume onto a known membrane area. The microscopy workflow examines small fields rather than the entire membrane. The calculation therefore depends on how representative the counted fields are of all captured particles. It also depends on retention efficiency and whether particles were lost to containers, filter apparatus or washing steps. The filter is not merely a passive sheet; its surface properties and loading influence the result.
Filters may have dense clusters near edges or patches where particles aggregate. If the analyst counts only the clearest-looking regions, the resulting average may not represent the membrane. A sampling plan should specify how fields are selected and how many are examined. Uniform distribution should be checked rather than assumed, especially for turbid or high-particle samples. A replicate filter prepared from the same water can help reveal preparation variability that repeated counting of one image would miss.
Overloaded filters are another failure mode. If bright points overlap, the microscope may treat multiple particles as one. Very sparse filters create the opposite problem: a few counted events produce large relative sampling uncertainty. The appropriate filtered volume and dilution depend on sample concentration and the validated method. Counting more fields cannot fully repair a preparation in which individual particles were never separated or retained reliably.
Staining changes visibility rather than creating biological identity
Nucleic-acid dyes increase the visibility of small particles by binding target material. This does not mean that every bright point contains an intact infectious viral genome. Free nucleic acids, aggregates and other stained material can generate background depending on preparation. The analyst uses morphology in the limited optical sense, fluorescence characteristics, blanks and validated counting thresholds to define the accepted events. The operational decision should be recorded so that another analyst can understand why an object was included or excluded.
Staining concentration, incubation timing, mounting method, light exposure and storage can influence fluorescence. The specific published protocol should govern preparation. Too little signal can create false negatives; excessive background can create false positives. Dye photobleaching and instrument sensitivity may alter a field as it is examined. A good quality-control plan monitors conditions without treating fluorescence brightness as a universal measure of viral genome size or infectivity.
An image showing fluorescent dots can be scientifically valuable without proving that every dot belongs to the same virus type. A species claim would require additional taxonomic or molecular evidence and its own uncertainty assessment. The reader should not allow a visually striking micrograph to replace the laboratory’s actual counting definition. The brightest image is not necessarily the most representative one.
From dots per field to particles per millilitre
The core calculation turns average accepted particle count per microscope field into an estimated total on the membrane, then divides by the original seawater volume represented. If the filter contains an effective number of microscope-field areas Nfields and the mean corrected count is c, estimated captured particles equal c times Nfields. The concentration equals that estimate divided by filtered volume. Apply dilution and recovery corrections only when their meaning and validation are known.
Suppose a hypothetical sample has thirty accepted VLPs per field after blank correction, the counted field geometry represents one thousand equivalent field areas per filter, and 0.5 mL of undiluted seawater passed through the filter. The estimated total is 30 × 1000 = 30,000 particles, giving 30,000/0.5 = 60,000 VLP per mL. If the sample had first been diluted tenfold and 0.5 mL of that diluted preparation was filtered, the original-water estimate would be 600,000 VLP per mL. Both the area ratio and dilution must be shown.
The result depends on effective filter area, microscope field area, particle distribution, capture efficiency and count definition. An arithmetic calculation is only one layer of validity. If the membrane has severe edge effects, multiplying one tiny field by a thousand can produce an implausible estimate. The analyst should verify that the area ratio was derived from instrument geometry and that counted fields were selected under a representative scheme.
Background correction has a biological meaning and a numerical rule
A field may contain fluorescent points unrelated to the particle population of interest. Appropriate staining and preparation blanks reveal some components of that background, although not every interfering object in a complex seawater matrix. A blank of three accepted-looking events per field and a sample mean of thirty-three may justify a corrected average of thirty when the laboratory has validated that subtraction. A blank measured with a different filter or imaging configuration may not represent the actual background. The correction belongs in the methods, not hidden in a spreadsheet cell.
Imagine the blank has substantial variation across fields. Subtracting an average of three from every sample field does not eliminate the uncertainty in that average. If a sample has only four events per field, a blank near three dominates the final inference and may make the estimate poorly resolved. A negative difference following blank subtraction does not mean seawater contains a negative concentration of viruses. It signals that the sample is indistinguishable from, or below, the method’s useful background resolution under those conditions.
Analysts should distinguish a procedural blank from a biological negative control. A filter processed using reagent or particle-free water tests staining and equipment background. A field matrix blank may test different contaminants. A known reference preparation can evaluate performance, but it does not automatically reveal every species of interfering particle in natural seawater. Document which control each row describes and how it contributes to accepted particle counts.
Dilution is a calculation, but it is also a preparation choice
Dilution may make a dense specimen countable by separating overlapping fluorescent events. If one volume of original seawater is combined with nine volumes of a suitable diluent, the total dilution factor is ten, provided the volumes and mixing are accurately controlled. A concentration calculated from the diluted preparation must then be multiplied by ten to estimate the original water. Forgetting that factor gives a tenfold undercount. Applying it twice gives a tenfold overcount.
However, the physical preparation can introduce losses or additional background. A suitable diluent and validated method must preserve particle recovery and staining behaviour. Dilution may also change aggregation, adsorption to tubes or interaction with the membrane. The factor of ten is not evidence that the particles were all retained. It corrects the represented original-water volume under an assumption that the procedure preserved the measurable population.
Consider two filters prepared from the same sample at different dilutions. If the corrected original-water estimates agree within justified uncertainty, that provides useful evidence of count linearity over the tested range. If the undiluted filter yields a lower estimate, particle overlap or background may be responsible. If the heavily diluted filter yields too few counts, random variation or adsorption may dominate. The disagreement is a diagnostic clue, not an instruction to choose whichever number appears most impressive.
Recovery efficiency can be measured but not assumed universal
Some particles may be lost during collection, transfer, filtration, staining and mounting. Recovery experiments can estimate the fraction that survives a specified process, often using appropriate reference material. If the measured estimate is 600,000 VLP/mL and a representative, validated whole-process recovery is eighty percent, a simple corrected estimate is 600,000 divided by 0.8, or 750,000 VLP/mL. That arithmetic is conditional on the recovery model applying to the actual population and processing chain.
Recoveries can vary with particle size, surface chemistry, concentration, matrix and handling. A single reference virus or synthetic particle does not necessarily behave like every natural virus-like particle. Applying a broad correction without validation can give a false appearance of accuracy. A good report distinguishes the observed recovered count from the model-corrected abundance and provides the assumptions and uncertainty in the recovery factor.
Recovery also interacts with blank correction. If an analyst subtracts background before correcting for recovery, they should be able to explain why that order matches the calibration design. If background fluorescence comes from a source not subject to the same recovery loss, multiplying the blank itself by a recovery factor can be inappropriate. This is a good example of how the order of arithmetic operations corresponds to real laboratory processes rather than being a matter of preference.
Counting uncertainty is more than a calculator’s last digit
At low event counts, counting statistics can be important. In a simple independent-event Poisson approximation, the standard deviation of a total count N is approximately the square root of N. If one hundred accepted events are counted in a fixed sampled area, the nominal counting standard deviation is about ten events, or ten percent relative. This approximation does not include uneven filter distribution, correlated clusters, staining variation or observer selection bias. It is a useful baseline, not a complete uncertainty budget.
Suppose two observers count the same fields and obtain values of one hundred and one hundred and twenty. Their disagreement may exceed what simple event counting statistics predict, depending on the procedure. They may have different thresholds for faint points or may split overlapping clusters differently. A blinded re-count, documented inclusion criteria and image analysis calibration can help separate observer variability from actual sample heterogeneity.
Repeatedly analysing the same image does not measure the full independence of field sampling. A stronger design includes multiple randomly selected fields, technical replicate filters and, when relevant, independently collected water samples. Each replication level answers a different question. Field-to-field variance describes spatial heterogeneity on the membrane; filter-to-filter variance reflects preparation; bottle-to-bottle variance includes sampling. Combining them without clear labels can hide the dominant uncertainty.
When bright objects overlap, the estimate can be biased downward
A heavily loaded field may show fluorescent clusters rather than distinguishable points. If each cluster contains multiple particles, counting clusters as individual events underestimates abundance. Software segmentation may attempt to separate them, but assumptions about particle size, intensity and optical blur influence the outcome. An algorithm that performs well on sparse images may fail on dense ones. A validation series across known dilutions is more informative than an arbitrary image-processing threshold.
The geometry of the filter can also become problematic. Particles near scratches, folds or dense debris may be invisible or excluded. A field that is difficult to count should be qualified according to a prespecified method rather than silently replaced with the clearest available area. Selective exclusion can bias a result if the difficult regions are systematically more particle-rich. A sample might require re-preparation at a lower loading density.
Photobleaching may reduce point brightness during repeated exposure. Camera gain, illumination and focus can change how many weak objects cross a detection threshold. Therefore an automated count is not completely instrument-independent. Retain imaging settings and test reproducibility on reference material. The correct interpretation is that the method counts events accepted under a specified detection rule, not that it reveals a metaphysically exact number of viruses in the original ocean.
Abundance versus infectivity, production and mortality
A high VLP count does not establish that the particles can infect a particular host. Infectivity depends on particle integrity, host susceptibility and biological conditions. Many aquatic viruses cannot be cultured readily, and fluorescent staining does not test whether they can initiate infection. The word ‘virus-like’ preserves this limitation. A community ecology study may use VLP abundance productively without claiming every counted event is capable of infection.
Likewise, abundance is not production rate. A population can be stable when viral production and loss are both high. Grazing, adsorption, sunlight, degradation and transport may remove particles while infections release new ones. An instantaneous count contains no direct time denominator. To estimate rates, researchers need an appropriate experimental or modelling design with repeated observations, controls and assumptions. Dividing a standing count by an arbitrary day does not magically create a measured rate.
Viral lysis of microbes can affect nutrient and carbon cycling, but a VLP count alone does not quantify the amount of carbon released. That inference needs information about hosts, infection dynamics, microbial biomass and other ecological processes. Responsible interpretation can discuss plausible connections and published mechanisms without equating every additional fluorescent dot with one cell lysed. The strength of the abundance measurement is precisely defined; other conclusions require additional evidence.
Viral-to-bacterial ratios are descriptive, not automatic mechanisms
A viral-to-bacterial ratio, often abbreviated VBR, compares counted virus-like particles with counted prokaryotic cells under specified methods and sampling conditions. If a hypothetical sample has 12 million VLP/mL and 1.5 million cells/mL, the ratio is eight. It is dimensionless because both denominators are millilitres. The calculation is easy, but the interpretation depends on comparable recovery and counting definitions for both populations.
A ratio of eight does not mean every cell has eight actively infecting viruses. VLPs may be free, inactive, adsorbed or unable to infect the sampled hosts. Cells in the denominator may include taxa with different vulnerability and metabolic activity. Different microscopy or flow-cytometry protocols may have distinct efficiencies for viral and cellular counts. The uncertainty in both numerator and denominator must be considered when comparing ratios across stations.
If a ratio rises, either the viral count increased, the cell count decreased, or both changed. A graph of the ratio alone cannot tell which. Read its components separately. The ratio can be a useful descriptive feature of an ecosystem, but it should not replace direct infection or production measurements when the scientific question concerns mechanisms of microbial mortality.
Field-of-view geometry is part of the denominator
The count under one microscope field must be related to the whole effective filter area before it becomes an estimate for the sample. If the effective membrane area is Afilter and each field covers Afield, the number of equivalent non-overlapping field areas is Afilter divided by Afield. Multiply that ratio by the representative mean corrected count per field and then divide by the filtered sample volume. An incorrect field diameter or magnification can create a systematic count error even when every visible particle was identified correctly.
Suppose the field area was measured using one objective but counts were made with another. The area ratio may change with magnification, so reusing the old ratio is unsafe. The numerical error may resemble a genuine change in viral abundance if different survey teams used different instruments. Retain a calibration record for field dimensions and the effective membrane area actually available for counting. The nominal membrane diameter may not equal the area occupied by uniformly filtered sample.
Field selection needs a rule that avoids preferential sampling of dense or sparse patches. Some protocols specify systematic traverses or random coordinates; what matters is that the method is documented and representative. If particle concentration varies strongly across the filter, additional fields or a new filter preparation may be necessary. The uncertainty is then larger than a simple Poisson count model. A statistically elegant formula cannot rescue biased selection of the images entered into it.
Worked example: combine the area ratio, blank and dilution
An invented filter has one thousand equivalent microscope fields in its effective area. Thirty-three accepted-looking fluorescent events are counted per sampled field on average, while a matched blank yields three. Under a validated blank subtraction the corrected mean is thirty. Multiplying by one thousand gives thirty thousand captured events. If the filtered preparation represents 0.5 mL of a tenfold-diluted seawater sample, the count is sixty thousand VLP per millilitre of diluted preparation, or six hundred thousand per millilitre of original water.
Now imagine the student forgets the dilution and reports 60,000 VLP/mL. The result is ten times too small although the microscopy counting itself was correct. Another student multiplies by both the inverse dilution and the nominal dilution volume separately, applying the factor twice and reporting six million. The repair is to write the amount represented by every denominator explicitly. The filtered 0.5 mL came from diluted water; the factor of ten restores the original-water basis only once.
Suppose a whole-process recovery study valid for the method shows eighty-percent recovery. The model-corrected estimate would be 600,000 divided by 0.8, or 750,000 VLP/mL. Report the observed recovered estimate and the corrected estimate separately. A correction introduces uncertainty from the recovery factor, and the study must justify its transfer to natural particles. A corrected count is not automatically more accurate when the recovery experiment used a nonrepresentative reference material.
Comparing microscopy with flow cytometry
Flow cytometry can measure fluorescent events as particles move through an optical interrogation zone. It offers a different route to estimating virus-like particle abundance, often with rapid data acquisition. However, cytometric gates, staining conditions, flow calibration, coincidence, background and event thresholds affect the accepted population. Agreement with epifluorescence microscopy must be evaluated using matched samples and a clear comparison design. The two methods do not automatically count exactly the same objects.
A cytometer’s fluorescence and scatter coordinates can separate operational particle populations, but they do not identify all virus species. A gate is a rule for including events in a measurement. Changing it can move the estimated abundance even if the seawater is unchanged. When comparing with microscopy, identify whether both assays include the same size and fluorescence range. Reference beads and flow calibration may assist instrument checks, but biological particle behaviour can differ from that of beads.
An important distinction is that microscopy can reveal spatial aggregates and filter-loading problems that may be hidden in a summary cytometric count, while cytometry can sample many events with a different measurement geometry. Each has strengths and limitations. A study that reports a method-specific difference should not immediately interpret the difference as biological change. The first question is whether the two analytical definitions are aligned.
Sequencing provides identities, not the same kind of abundance
Marine viral metagenomics and targeted molecular methods can reveal genetic diversity and sequences associated with viruses. Their outputs are not automatically equivalent to microscopy VLP per millilitre. Read counts depend on extraction yield, genome size, amplification where used, library preparation, sequencing depth and bioinformatic decisions. A taxonomic assignment is based on sequence evidence and database methods, not on the brightness of one fluorescent point.
A sample may have many microscopic VLPs but yield relatively few confidently classified viral sequences because genomes are unknown, nucleic-acid recovery is uneven or the chosen pipeline discards uncertain reads. Conversely, many reads assigned to a viral sequence do not necessarily represent the same number of intact particles in the original water. A molecular result may be highly informative about community composition while remaining unsuitable as a direct total particle count without calibration.
The most productive synthesis explains what each method contributes. Microscopy estimates an operational abundance of stained particle-like events. Flow cytometry estimates gated fluorescent events in a calibrated flow system. Sequencing helps characterise genetic material and possible identities. Culture or infection assays test different biological properties. They can complement one another, but a report should never replace one output label with another simply because all are related to marine viruses.
Time series: abundance changes do not equal lysis rates
Suppose a water sample shows eight million VLP/mL on Monday and ten million on Tuesday. The difference is two million per millilitre across the measured interval. Without matched sampling, transport context and appropriate replication, the difference cannot be called a production rate. Even within the same parcel, appearance, decay, aggregation, settling and exchange can alter counts. The observed net change is a starting point for process analysis, not a complete carbon-cycle budget.
A carefully designed experiment may measure particle appearance or loss across multiple controlled timepoints, with appropriate corrections for dilution and handling. Yet a net increase still need not equal total viral production if losses occur simultaneously. A lysis estimate involves host processes and usually needs additional biological evidence. Reporting a standing abundance per hour by simply dividing by elapsed time is invalid; it assumes the initial pool was newly produced during that time.
Temporal sampling also requires care. A station revisited after a tide change may contain a different water mass. If its VLP count changes, advection can be involved even when the geographic coordinates match. Temperature, salinity and current context help determine comparability. The correct interpretation separates a time series of samples at a location from a time series following the same water parcel.
A realistic uncertainty budget has several levels
Counting uncertainty arises from the finite number of accepted events. Spatial heterogeneity arises from variation among microscope fields and regions of the membrane. Preparation uncertainty includes filtration, dilution, staining, particle loss and recovery. Sampling uncertainty concerns whether the bottle represented the intended water mass. These uncertainties do not necessarily behave independently. A single number after a plus-minus sign should identify which components it represents.
Consider one hundred accepted particles on a set of randomly chosen fields. Under the simplest independent-event model, square root of one hundred gives a count standard deviation near ten, or ten percent relative. If fields contain clustered particles, the actual variation may be much higher. If the same faulty dilution factor was used throughout, replicating fields will not remove the systematic tenfold error. Statistical replication and procedural verification solve different problems.
Uncertainty in a ratio needs attention too. If viral abundance is 12 million per millilitre and bacterial abundance 1.5 million per millilitre, VBR equals eight. If both counts have independent uncertainty, the ratio uncertainty reflects both. If the same sample volume error affects numerator and denominator proportionally, some error may cancel. The analyst should not automatically add or subtract uncertainty percentages without considering correlation and the defined counting procedures.
Independent problems for a reader to solve
Problem one: a filter represents 800 microscope fields, average sample count is forty events per field, and a matched blank is five. The corrected mean is thirty-five, giving 28,000 estimated particles on the filter. If 0.2 mL of undiluted water was filtered, concentration is 140,000 VLP/mL. Problem two: the preparation was actually diluted fivefold first. The original-water estimate becomes 700,000 VLP/mL. In both cases the answer follows from a declared area ratio, filtered volume and dilution factor.
Problem three: a measured estimate is 800,000 VLP/mL and a validated recovery factor is 0.8. The corrected estimate is one million VLP/mL. If recovery uncertainty is large, report that rather than presenting the result as exact. Problem four: two observers count the same slide but use different fluorescence thresholds. Their different results do not immediately prove ecological heterogeneity; they first reveal an analytical definition issue. A common threshold protocol and blinded comparison would help test whether the difference persists.
Problem five: viral abundance is 6 × 10^6 VLP/mL and a paired cellular count is 2 × 10^6 cells/mL. The ratio is three, but it does not mean each cell has three active infections. Problem six: an instrument reports a ten-percent increase in VLP abundance after one hour. The statement ‘lysis released ten percent more viruses’ is unsupported without host and process controls. A reader should be able to give the correct arithmetic and then state precisely where biological interpretation requires additional data.
Frequently asked questions that clarify the limits
Are fluorescent VLPs confirmed infectious viruses? No. The assay establishes stained objects meeting an operational counting criterion. Are all virus-like particles the same size or taxonomic group? No, and ordinary fluorescence microscopy generally cannot resolve their detailed identities. Does a higher count prove more viruses were produced during the sampling hour? No. It could reflect transport, losses, concentration of water masses or preparation differences as well as production.
Can filtering more water always improve accuracy? No. Overloading may cause overlapping particles, clogging and nonuniform distribution. Can a flow cytometer count be compared with a microscope count? Yes, with clear matched methods and intercomparison, but the underlying event definitions must be checked. Does sequencing read abundance equal VLP abundance? No. Sequences and fluorescent particle events have different denominators and biases.
Is a corrected count better than a raw count? It may be, if blank and recovery corrections are validated and clearly reported. A poorly justified correction can introduce error. Does an elevated viral-to-bacterial ratio measure infection intensity? Not directly. It is a descriptive ratio of two measured standing inventories. Its interpretation improves when coupled with host ecology, process assays, physical context and explicit uncertainty.
The reader’s final audit before accepting a viral abundance claim
Start with the water sample: when and where it was collected, how it was handled and what original volume was represented. Move to the filter: membrane type, pore size, effective area, preparation, dilution and loading pattern. Move to the microscope or image system: stain, optical settings, accepted event criteria, field selection and blank. Then inspect the calculation from events per field to original-water VLP per millilitre. Each denominator should be explicit and capable of independent checking.
Next examine controls and uncertainty. Were blank counts substantial? Was recovery measured under comparable conditions? Did replicate filters agree? Were enough fields examined to reveal spatial heterogeneity? Did observers or image software use compatible thresholds? If an answer is missing, the count may still convey information, but a reader should not assign it the certainty of a fully audited abundance estimate. An uncertainty statement should not conceal an unknown systematic loss.
Finally read the biological interpretation for overreach. A VLP concentration is neither an infection rate nor an ecosystem-wide lysis flux. A ratio is not a direct host assignment. A molecular identity is not the same as a fluorescent point count. Good scientific explanation preserves the observed quantity, shows the arithmetic and clearly marks each step beyond it. This is how a complex but useful marine microbiology method can be made accessible without changing what the evidence actually says.
Original methodological sources and related reading
PubMed record for the SYBR Gold marine viral particle enumeration study
eduKateSG How to Translate Easily series index
Related eduKateSG guide: marine phytoplankton flow-cytometry gates and cell counts
Related eduKateSG guide: marine eDNA metabarcoding, blanks and sequence reads