A glucose strip, a pregnancy test and a laboratory surface plasmon resonance instrument look nothing alike, yet they share one idea. A biological molecule that has evolved, or been selected, to recognize one partner out of thousands is placed next to a physical device that can report what happened. The recognition event is chemistry; the report is a current, a voltage, a change in colour, a frequency shift or a few millikelvin of warming. A biosensor is the engineered bridge between the two.
This chapter builds the vocabulary and the mental model that the rest of the book relies on. It explains what the WIA-BIO-011 standard means by a biosensor, retraces briefly how the field arrived at today's devices, and walks through the architectural layers that sit between a drop of sample and a number on a screen. Most of the chapter is about the recognition layer: enzymes, antibodies, aptamers, nucleic-acid probes and a few less common elements, together with the practical problem of fixing them to a surface without destroying them. The physics of transduction waits for Chapter 2 and the formal performance metrics wait for Chapter 3; where this chapter needs a number, it uses simple, checkable arithmetic.
The intended reader is an R&D engineer, a point-of-care product developer, a firmware or data engineer who connects readers to apps, or a quality and regulatory specialist. First-year chemistry (concentration, redox, pH) and basic statistics are enough. No electrochemistry or optics background is assumed.
Section 1.4 of the specification defines a biosensor as an analytical device that combines a biological recognition element with a transducer. Four terms carry that definition, and it is worth fixing them before going further.
The key word in the definition is combines. The recognition element is not a reagent that is added in a separate step and then read on another instrument. It is held on, or in intimate contact with, the transducer, so that the recognition event and the physical signal happen in the same place. That integration is what allows a biosensor to be small, fast and, in the case of continuous monitors, to report a changing concentration without anyone adding reagents.
Section 2.1 of WIA-BIO-011 draws the device as a linear chain:
Sample → Biorecognition Element → Transducer → Signal Processor → OutputThe chain is a useful simplification, and it is also the shape of the data that the standard's SDK records. A BiosensorConfig object names the transducer (type, with an electrochemical or optical subtype), the recognition element (biorecognitionElement) and the analyte (targetAnalyte); a calibration record describes the signal processor's mapping; a measurement record is the output. Chapter 4 describes those records field by field. For now the point is that every box in the chain becomes something an implementation must be able to describe, not only something an engineer must build.
Real devices bend the chain in two ways. First, information flows backwards: the reader sets the electrode potential, the excitation wavelength or the temperature, and those settings change what the recognition layer does. Second, the sample is not passive. Blood, urine, saliva, food extracts and environmental water carry proteins that foul surfaces, redox-active molecules that produce current on their own, and salts and pH shifts that change how biomolecules fold. Much of biosensor engineering, and much of Chapter 6, is about those back-channels.
Two neighbouring categories are easily confused with biosensors, and the differences matter when an engineer decides which requirements apply.
A laboratory analyzer in a central laboratory may use exactly the same enzymes and antibodies, but it dispenses them as liquid reagents, runs them through incubation and washing stations, and reads them with a separate photometer or luminometer. It trades integration for throughput and flexibility. Its calibration is managed by trained staff with commercial calibrators; the biosensor, by contrast, often has to calibrate itself or rely on factory calibration of a manufacturing lot.
A chemical sensor has a transducer and a selective layer, but the selective layer is not biological. A glass pH electrode, an ion-selective electrode for potassium or an oxygen electrode are chemical sensors. The distinction is not academic: the same potentiometric transducer becomes a biosensor when a layer of urease is placed on top of a pH-sensitive surface, because the enzyme now decides which molecule produces the pH change. The standard's AnalyteCategory enum still includes ion, because ions are frequently measured by biosensor platforms and appear in the same data flows.
The specification's reference list opens with Clark and Lyons (1962), "Electrode systems for continuous monitoring in cardiovascular surgery". Clark had already built an oxygen electrode protected by a gas-permeable membrane. The step that made biosensors possible was to trap glucose oxidase next to that electrode. Glucose oxidase consumes oxygen as it oxidizes glucose, so the fall in oxygen current became a measure of glucose. That single idea — an enzyme held against an electrochemical transducer — is still the template for most commercial biosensors.
Glucose sensors have since been described in three "generations", a classification based on how electrons travel from the enzyme to the electrode.
Two later developments define much of today's landscape. Continuous glucose monitors (CGM) place a small enzyme electrode under the skin, where it measures glucose in interstitial fluid for days at a time. Lab-on-chip devices integrate sample handling and detection in microfluidic channels, which is the "miniaturization" theme of spec §2.2. Alongside these, affinity biosensors based on antibodies and nucleic acids have grown from laboratory instruments into lateral-flow strips and portable readers. The review by Turner (2013), "Biosensors: sense and sensibility", and the survey of commercial applications by Bahadır and Sezgintürk (2016) are the other two references the specification cites; both are good entry points to the literature. This book does not give market figures: they change quickly and are not part of the standard.
Spec §2.1.1–§2.1.4 break the chain into four layers and give indicative ranges for each. Figure 1-1 puts those ranges on the diagram.
The sample interface delivers the sample to the recognition layer and, where needed, prepares it. The specification lists flow cells, microfluidics and direct immersion as delivery modes; filtration, dilution and extraction as preparation steps; and temperature, pH and ionic strength as the environmental variables to control. The microfluidic volume range of 1–100 µL (§2.2) is a reminder of how little sample a modern device may see. A fingerstick meter works with a fraction of a microlitre, so every molecule that sticks to a channel wall or evaporates from a drop is a larger fraction of the total than in a laboratory cuvette.
The recognition layer holds the biological element in an immobilization matrix. The specification names four matrix families: membranes, polymers, sol-gels and self-assembled monolayers (SAMs). The quoted thickness range, 10 nm to 100 µm, spans four orders of magnitude, and the two ends behave very differently. A SAM of a few nanometres on gold places the recognition molecules inside the region that surface-sensitive transducers such as SPR can probe. A hydrogel or polymer film tens of micrometres thick can hold far more enzyme, but analyte and products must diffuse through it, which slows the response and, as Section 1.4 shows, reshapes the calibration curve. The binding capacity range of 10¹² to 10¹⁵ molecules per cm² reflects the same spread: the low end is roughly what a single packed layer of large proteins provides, the high end is closer to a dense layer of small molecules or a three-dimensional matrix.
The transducer is an electrode for electrochemical sensors, a waveguide or optical surface for optical sensors, or a crystal for piezoelectric sensors. The quoted active area of 0.01 to 100 mm² ranges from microelectrodes to large screen-printed pads. Area matters because many signals scale with it: an amperometric current is proportional to electrode area, which is why Chapter 3 recommends area-normalized sensitivity when comparing electrodes.
The processor amplifies (gains of 10³ to 10⁶), filters (bandwidths of 0.1 to 100 Hz) and digitizes (12 to 24 bit). A short calculation shows why those numbers belong together. Suppose, as an example, an amperometric front end converts current to voltage with a gain of 10⁶ V/A, so 1 µA becomes 1 V. A 16-bit converter spanning ±1 V has a step of 2 V / 65,536 ≈ 30.5 µV, which corresponds to about 30 pA of electrode current. That is ample for a glucose strip that produces microamperes, but it may be marginal for a microelectrode whose entire signal is a few nanoamperes, which is one reason higher-resolution converters and narrower bandwidths appear in low-level sensors. The final stage of the processor is the calibration model; everything about building and checking it is the subject of Chapters 3 and 6.
| Layer (spec) | Indicative range | Typical error it introduces | Where the book returns to it |
|---|---|---|---|
| Sample interface (§2.1.1, §2.2) | 1–100 µL microfluidic volume | Matrix effects, incomplete filling, evaporation, temperature drift | Ch6 (interference), Ch7 (field devices) |
| Recognition layer (§2.1.2) | 10 nm–100 µm thick; 10¹²–10¹⁵ sites/cm² | Loss of activity, leaching, fouling, slow diffusion | This chapter; Ch3 (stability) |
| Transducer (§2.1.3) | 0.01–100 mm² active area | Background current, reference drift, surface ageing | Ch2 |
| Signal processor (§2.1.4) | Gain 10³–10⁶; 0.1–100 Hz; 12–24 bit | Quantization, noise aliasing, wrong calibration model | Ch3, Ch5 |
Enzymes were the first recognition elements and are still the most widely used. They are catalysts: one enzyme molecule converts many analyte molecules, so the signal is amplified at the recognition step itself. The specification tabulates four canonical examples.
| Enzyme | Analyte | Reaction | What is detected | EnzymeType value |
|---|---|---|---|---|
| Glucose oxidase (GOx) | Glucose | Glucose + O₂ → gluconic acid + H₂O₂ | Oxidation of H₂O₂ at the electrode | glucose-oxidase |
| Lactate oxidase (LOx) | Lactate | Lactate + O₂ → pyruvate + H₂O₂ | Oxidation of H₂O₂ | lactate-oxidase |
| Urease | Urea | Urea + H₂O → NH₃ + CO₂ | Local pH change | urease |
| Acetylcholinesterase (AChE) | Pesticides (as inhibitors) | Enzyme activity is inhibited | Decrease in current | acetylcholinesterase |
The EnzymeType enum in the SDK adds three more values: cholesterol-oxidase, a third oxidase that produces peroxide; and horseradish-peroxidase and alkaline-phosphatase, which are used less as recognition elements than as labels in immunoassays, where they turn each bound antibody into a source of many detectable product molecules.
The rate of an enzyme reaction with a single substrate follows, to a good approximation, the Michaelis–Menten equation given in spec §3.1.2:
v = Vmax · [S] / (Km + [S])Here v is the reaction rate, Vmax the rate when every enzyme molecule is busy, [S] the substrate concentration and Km the Michaelis constant — the substrate concentration at which the rate is half of Vmax. For a biosensor the measured current follows the same shape, but with an apparent constant:
I = Imax · [S] / (Km,app + [S])The constant is "apparent" because the sensor does not see the enzyme's intrinsic kinetics. The substrate must reach the enzyme by diffusing through the sample, a membrane and the immobilization matrix, and the co-substrate (oxygen, or a mediator) must be present in sufficient supply. Whatever limits the flow of substrate to the enzyme changes the constant the sensor exhibits from outside.
Two limits of the equation explain most of what an engineer needs. When [S] is much smaller than Km,app, the denominator is essentially Km,app and the current is proportional to concentration: I ≈ (Imax / Km,app) · [S]. That is the linear region used for calibration. When [S] is much larger than Km,app, the current approaches Imax and stops depending on concentration: the sensor saturates. The ratio between the true current and the straight-line extrapolation of the initial slope is simply Km,app / (Km,app + [S]). The linear range therefore sits well below Km,app, not up to it.
The SDK stores a default Michaelis constant for glucose oxidase, BIOSENSOR_CONSTANTS.GOX_KM = 33.0 mM. It is a default, not a constant of nature: literature values vary with the enzyme source, the assay conditions and, above all, with immobilization. It is nevertheless a convenient number to test the question every glucose-sensor designer faces: can an enzyme with this constant give a linear response over a clinical range that reaches about 30 mM?
| [S] (mM) | I / Imax = [S]/(33 + [S]) | Linear extrapolation [S]/33 | Shortfall from linear |
|---|---|---|---|
| 1 | 0.029 | 0.030 | −2.9 % |
| 5 | 0.132 | 0.152 | −13.2 % |
| 10 | 0.233 | 0.303 | −23.3 % |
| 20 | 0.377 | 0.606 | −37.7 % |
| 30 | 0.476 | 0.909 | −47.6 % |
The answer is clearly no. At 30 mM the response is almost half what a straight line through the low-concentration points predicts. If the criterion is that the response may fall at most 5 % below linear, the condition Km,app / (Km,app + [S]) ≥ 0.95 gives [S] ≤ Km,app / 19, which for 33 mM means about 1.7 mM. Turning the inequality around, a 5 %-linear response up to 30 mM requires an apparent constant of at least 19 × 30 = 570 mM. These numbers are illustrative — Chapter 3 treats linearity criteria properly — but the order of magnitude is the lesson: a practical glucose electrode must make the enzyme appear far less sensitive to glucose than it really is.
The standard solution is an outer membrane that limits how fast glucose reaches the enzyme while letting oxygen through relatively easily. With the membrane in place, the enzyme layer always sees a much lower glucose concentration than the sample, so it stays in its own linear region. The current is then set mainly by the diffusional flux across the membrane, and diffusional flux is proportional to the concentration difference that drives it. Seen from outside, the sensor behaves as if its Michaelis constant were much larger: Km,app rises and the linear range extends accordingly. The price is a smaller current and a slower response, because the membrane that holds back glucose also slows its arrival. The same membrane also protects the enzyme from proteins in the sample and, in first-generation designs, eases the oxygen shortage by admitting relatively more oxygen than glucose.
Acetylcholinesterase shows a different way of using an enzyme. Organophosphate and carbamate pesticides inhibit the enzyme, so the sensor does not measure a substrate: it measures how much activity is lost. The usual procedure is to record the enzyme's response to a fixed amount of substrate before (I₀) and after (Iᵢ) exposure to the sample and to report the percentage inhibition 100 · (I₀ − Iᵢ) / I₀. Three consequences follow. The signal decreases as the analyte increases, so the calibration slope is negative; the result depends on incubation time, because inhibition builds up; and the sensor may be irreversibly consumed, which fits a single-use format. Inhibition-mode sensors report a class of compounds rather than one molecule, because many inhibitors act on the same enzyme. Chapter 7 returns to them for environmental monitoring.
Antibodies do not convert the analyte; they bind it. An immunosensor therefore needs a way to make binding visible. Spec §3.2.1 names three configurations, and the choice among them depends mainly on the size of the analyte.
Analyte + Antibody ⇌ Complex). The transducer registers the binding itself, with no label: the added mass on a crystal, the change of refractive index at an SPR surface, or the change in charge-transfer resistance measured by impedance. Direct detection is simple and works in real time, but the signal from a small molecule is tiny, and anything else that sticks to the surface produces the same kind of signal.The strength of binding is expressed by the association constant Ka = [Ab–Ag] / ([Ab] · [Ag]), in M⁻¹, or more commonly by its reciprocal, the dissociation constant Kd = 1 / Ka, in molar units. A lower Kd means tighter binding. The specification gives two bands: high affinity with Kd from 10⁻¹² to 10⁻⁹ M (picomolar to nanomolar), and medium affinity from 10⁻⁹ to 10⁻⁶ M (nanomolar to micromolar).
Why does Kd set the floor of what a sensor can measure? At equilibrium, and when the analyte is not depleted by binding, the fraction of antibody sites occupied is θ = [Ag] / (Kd + [Ag]). Table 1-4 applies this to an antibody with Kd = 1 nM.
| Analyte concentration | [Ag] / Kd | Fraction of sites occupied θ |
|---|---|---|
| 10 pM | 0.01 | ≈ 1.0 % |
| 100 pM | 0.1 | ≈ 9.1 % |
| 1 nM | 1 | 50 % |
| 10 nM | 10 | ≈ 91 % |
Two orders of magnitude below Kd, only about one site in a hundred is filled. A label-free transducer then has to distinguish that one per cent from background and from nonspecific adsorption, which is often impossible. Measuring well below Kd requires either a higher-affinity antibody or a label that turns each bound molecule into many detectable events — which is exactly why enzyme labels such as horseradish peroxidase and alkaline phosphatase appear in the SDK's enzyme list.
Kinetics tighten the constraint further. Kd is the ratio of the dissociation and association rate constants, koff / kon, and the time constant for approaching equilibrium is about 1 / (kon · [Ag] + koff). Take, as an illustrative example, kon = 10⁵ M⁻¹·s⁻¹ and Kd = 1 nM, so koff = 10⁻⁴ s⁻¹. At 10 pM the time constant is roughly 1 / (10⁻⁶ + 10⁻⁴) s ≈ 9,900 s, close to three hours. A ten-minute point-of-care test therefore reads the sensor long before equilibrium, and its calibration is a calibration of a kinetic, time-dependent signal. That is why read time is a fixed part of an immunoassay protocol.
Aptamers are short single-stranded DNA or RNA molecules that fold into a shape that binds a particular target. They are found by iterative selection from very large random libraries rather than raised in animals. Spec §3.3 lists their advantages and disadvantages fairly. They are made by chemical synthesis, so they are reproducible from batch to batch and easy to modify with thiol, amine or redox-label groups at a chosen position. They tolerate heat and can often be regenerated by unfolding and refolding, and they can target anything from small molecules to proteins and whole cells. On the other side, nucleases in biological samples degrade them unless the backbone is chemically protected, and because their binding depends on folding, their affinity changes with ionic strength, divalent cations and pH. That last point is more than a footnote: a sensor calibrated in buffer can read differently in urine or serum simply because the salt content differs, which is a matrix effect discussed in Chapter 6.
Nucleic-acid probes (dna and rna in the enum) detect a target sequence by base pairing: probe + target ⇌ duplex. The stability of the duplex is summarized by its melting temperature Tm, the temperature at which half of the duplexes have come apart. Spec §3.4.1 quotes a widely used empirical estimate:
Tm ≈ 81.5 + 0.41 · (%GC) − 675 / N (°C; N = probe length in nucleotides)Two cautions apply. The formula is an empirical rule meant for probes of roughly 14 to 70 nucleotides, not a thermodynamic model. And this version omits the salt term: the fuller form adds 16.6 · log₁₀[Na⁺], which is zero only at 1 M sodium. For a 20-nucleotide probe with 50 % GC the formula as quoted gives 81.5 + 20.5 − 33.75 ≈ 68 °C; at 50 mM sodium the salt term adds 16.6 × log₁₀(0.05) ≈ −21.6 °C, bringing the estimate to about 47 °C. A difference of that size decides whether a single-base mismatch can be discriminated at the assay temperature. For design work, nearest-neighbour thermodynamic tools are the better choice; the simple formula is useful as a sanity check.
The BiorecognitionElement enum contains three further values that the specification's prose does not discuss in detail.
whole-cell sensors use living microorganisms or cells as the recognition element. They respond to a class of effects — toxicity, bioavailability of a metal, an inducer of a reporter gene — rather than to a single molecule. They are slow and need care to keep alive, but they answer questions that affinity sensors cannot, such as whether a sample is biologically harmful.bacteriophage elements use viruses that infect particular bacteria. Their tail proteins recognize the host strain, and phages tolerate conditions that would damage antibodies. They are used mainly for bacterial detection in food and water.molecularly-imprinted-polymer (MIP) is a synthetic polymer cast around a template molecule that is then removed, leaving shaped cavities. MIPs are not biological, but they play the same role and are recorded in the same field. They are robust and cheap; selectivity and batch-to-batch consistency are their usual weaknesses.Table 1-5 summarizes the options. The stability and cost columns are qualitative comparisons for orientation, not measured values.
| Element | Typical analytes | Key affinity / kinetic parameter | Stability | Relative cost | Enum value |
|---|---|---|---|---|---|
| Enzyme | Metabolites (glucose, lactate, urea, cholesterol); inhibitors such as pesticides | Km,app; turnover; inhibition % | Moderate; sensitive to temperature and drying | Low to moderate | enzyme or an EnzymeType value |
| Antibody | Proteins, hormones, drugs, toxins, pathogens | Kd (or Ka); kon, koff | Moderate; denatures if mishandled | Moderate to high | antibody |
| Aptamer | Small molecules, proteins, cells | Kd; dependence on ions and pH | High thermally; nuclease-sensitive | Low once selected | aptamer |
| DNA / RNA probe | Nucleic-acid sequences | Tm; mismatch discrimination | DNA high; RNA lower | Low | dna, rna |
| Whole cell | Toxicity, bioavailable metals, reporter inducers | Dose–response of the cell | Low; living system | Low materials, high upkeep | whole-cell |
| Bacteriophage | Specific bacteria | Host range; capture efficiency | High | Low to moderate | bacteriophage |
| Molecularly imprinted polymer | Small molecules, some proteins | Imprinting factor; binding isotherm | Very high | Low | molecularly-imprinted-polymer |
In the SDK the choice is recorded in BiosensorConfig. The field biorecognitionElement accepts a BiorecognitionElement value, an EnzymeType value or any string, and it is one of the three fields that the SDK's configuration check requires, together with type and targetAnalyte. A first-generation glucose electrode could be described like this (the potential is an example value):
import type { BiosensorConfig } from '@wia/bio-011';
const gluConfig: BiosensorConfig = {
type: 'electrochemical', // required
electrochemicalType: 'amperometric',
biorecognitionElement: 'glucose-oxidase', // required: an EnzymeType value
targetAnalyte: 'glucose', // required
analyteCategory: 'metabolite',
workingElectrode: 'platinum',
referenceElectrode: 'ag-agcl',
counterElectrode: 'platinum',
appliedPotential: 0.6, // V vs Ag/AgCl, example value
sampleType: 'blood',
temperature: 25, // °C
pH: 7.0
};Chapter 2 explains why a peroxide-detecting electrode sits at around this potential and what it costs in interference; Chapter 4 describes every field of the configuration.
A recognition element that works perfectly in solution can lose most of its activity once it is fixed to a surface. Immobilization methods — physical adsorption, entrapment in a polymer or sol-gel, covalent coupling, cross-linking, affinity capture such as biotin–streptavidin, and thiol SAMs on gold — each trade ease against control. Four failure modes recur, and each one surfaces later in the book as a performance problem.
An antibody binds its target with the tips of its two arms. Attached at random, a large share of antibodies face the surface with their binding sites blocked. Oriented coupling, for example through the antibody's constant region using protein A or G or through site-specific chemistry, increases the fraction of active sites. The same issue applies to aptamers, which should be attached at one end, and to enzymes whose active site lies near the attachment point. Poor orientation shows up as low sensitivity and a poor detection limit, the metrics of Chapter 3.
Proteins unfold on hydrophobic surfaces, during drying, under reactive coupling chemistry and at elevated temperature. Dry-reagent formats such as test strips are especially exposed, because the enzyme must survive drying during manufacture and rehydration during use. Stabilizers such as sugars and polymers are added for this reason. Denaturation often happens gradually in storage, so it appears as a shelf-stability problem rather than an immediate failure.
Physically adsorbed or loosely entrapped elements, and soluble mediators, can slowly escape into the sample. Each measurement then starts with a little less recognition element than the one before. In a continuous sensor this produces a steady fall in sensitivity — the drift and operational stability discussed in Chapter 3 — and in an implanted device it raises questions about what is released into the body. Covalent coupling and cross-linking reduce leaching at the cost of some activity.
Samples are full of molecules that stick to surfaces. Proteins in blood can coat an electrode within minutes; on a label-free affinity sensor, that coating produces a signal indistinguishable from real binding. Blocking agents (bovine serum albumin, casein), antifouling layers (polyethylene glycol, zwitterionic polymers) and permselective membranes limit the problem. It is also the physical origin of many of the interferences that Chapter 6 teaches readers to test for, and it is why a sensor calibrated in buffer must be verified in the matrix it will actually measure.
A useful habit is to write down, for any new design, which of the four failure modes the chosen immobilization is most exposed to, and to plan an experiment that would reveal it early. The ENZYME_TEMP_COEFFICIENT of 3 %/°C that the SDK uses for temperature correction is a good example of an assumption that depends on the immobilized state: the default may be reasonable for one enzyme layer and wrong for another, so Chapters 3 and 6 treat it as something to measure.
WIA-BIO-011 consists of a main specification (spec/WIA-BIO-011-v1.0.md, sections 1 to 10 and Appendix A), a TypeScript SDK (@wia/bio-011, with types.ts defining the data model and index.ts implementing BiosensorSDK), and a command-line tool (wia-bio-011). Table 1-6 shows where each chapter draws its material.
| Chapter | Topic | Spec sections | SDK / CLI artifacts |
|---|---|---|---|
| 1 | Fundamentals and recognition elements | §1–§3, §10.1 | BiorecognitionElement, EnzymeType, AnalyteCategory |
| 2 | Transduction mechanisms | §4 | BiosensorType, ElectrochemicalType, OpticalType, BIOSENSOR_CONSTANTS |
| 3 | Signal processing and performance metrics | §5, §6, Appendix A | Signal-processing, sensitivity and LOD helpers |
| 4 | Data model | (defined in code) | BiosensorConfig, CalibrationData, MeasurementResult, QualityFlag |
| 5 | API, SDK and CLI | (defined in code) | BiosensorSDK, error codes, wia-bio-011 |
| 6 | Calibration, measurement and QC protocol | §7, §5.4 | calibrateSensor, measureAnalyte, runQualityControl |
| 7 | Point-of-care, wearable and environmental integration | §8, §10.3, §10.4 | POCTestResult, TimeSeries, BatchMeasurement |
| 8 | Validation, conformance, certification and outlook | §9, §10.2 | validateSensor |
Readers with different jobs can take different routes through the book.
Version 1.0 is a useful but imperfect document, and the book says so where it matters. Three points are relevant already. First, the glossary in §1.4 defines "dynamic range" as the range over which the response is linear. That is the linear range; the dynamic range is the wider span from the detection limit to the highest measurable concentration, and Chapter 3 keeps the two apart. Second, the reference list in §10.1 is numbered 1, 3 and 5; only those three works exist in the document, and this book cites no others as if they were part of it. Third, the repository also contains four files named PHASE-1 to PHASE-4. They carry a different standard identifier, and their headings do not match their file names; the book uses a few general statements from them only as clearly labelled, non-normative context. Where the specification's numbers or the SDK's behaviour need correcting, the text says so on the spot and Chapter 8 collects every item in one errata table.
I = Imax[S]/(Km,app + [S]). The linear range lies well below Km,app: with the SDK default of 33 mM the response at 30 mM is 48 % below linear, which is why diffusion-limiting membranes are used to raise the apparent constant.Kd sets the practical floor: two decades below Kd only about 1 % of sites are occupied, and at low concentrations equilibrium can take hours.BiorecognitionElement vocabulary; the quoted Tm formula omits salt and applies only to probes of roughly 14–70 nucleotides.