Published in



The Wearable Biosensor Opportunity

A Unified Design of Consumer Electronics and Biological Assays

What do you get when you cross-fertilize a Fitbit™ step counter with ELISA? A wearable biosensor.

I expect that for most readers, the stuff of biological engineering — DNA polymers, proteins, monosaccharides, cells — are rarely thought of in combination with the stuff of electronics engineering — capacitors, serial ports, MOSFETs, accelerometers. And for good reasons.

One is condensed — or wet — matter of mostly carbon and elements in the vicinity of carbon in the periodic table. The other is solid-state matter of a wide range of elements, especially from the third period and below in the periodic table, as well as plastics for structural scaffolding. One is functionally regulated by specific binding events to one or more types of molecules, or ligands, at varying concentrations and conditions of the milieu (e.g. pH and temperature). The other is regulated above all else by the current and voltage of electricity.

This is a challenging picture.

If the combined work in these two engineering domains is to produce novel and useful designs, the structural-functional chasm must be bridged so that one can predictably be altered by a change in the other. Then add to that the effort to creatively overcome institutional friction, which favours established educational and professional specializations.

There have nonetheless been innovations and commercializations in this mixture of engineering and knowledge domains. Biosensors is a class of applications that depends on a working solution to at least some aspects of the outlined challenges. The conventional method of diabetes monitoring, as we will see later, is an excellent example.

Diabetes self-monitoring — one of the more common applications of biosensors.

A further specification is a wearable biosensor, that is, a consumer device to be worn on the body, or at least consumer-grade and mobile. That additional specification constrains the product requirements further: small, sturdy, painless, frictionless user onboarding, accurate outside the controlled lab environment, and certifiable for global consumer market access. It is the figurative nuts and bolts of this wearable class of consumer products that I look closer into.

And that brings us back to the imagined cross-fertilized product in the first paragraph. There are useful parallels and contrasts to be found in the engineering details of the archetypical step counter on the wrist and the commonplace biological assay in the lab.

With that in mind, I begin with a “teardown”, down to the very atoms, so to speak, of the two parent devices before I delve deeper into the biosensor opportunity.

From an Arm in Motion to Cheerful Push Notification via the Properties of Atoms

Consider the elementary function of a consumer wearable device: step counting. Attached to the wrist, the step counter device senses distinct body movements and converts that into step data, which we represent digitally and store in databases, maybe for our health tracking app to later tell us about trends in our mobility.

Body movement implies mechanical force. Recall that any acceleration or deceleration of an object comes about through a force acting on the object. So the motion back and forth of the arm when on a stroll, the impulse through the body when the foot hits the ground, and the torque through the hip joints as the body is propelled forward, all involve mechanical forces of various directions, magnitudes, and origins.

These forces are the basic physical facts to the application. They are at the foundation of the utility that a step counter gives the user. But to go from the former to the latter, a transformative process is needed.

The very first unit in that long chain of processes is the sensing element. In many step counters, the sensing element is composed of a seismic mass and a deformable structure, like a cantilever beam. So what is that?

Cartoon illustration of a seismic mass attached to a deformable cantilever beam. In the top-most illustration, no force is applied to the seismic mass; the bottom-most illustration is when force bends the cantilever beam.

The image above is a cartoon illustration of the design. A thin projecting cantilever beam is attached at one end to a block of known mass, called the seismic mass. A thin layer of a second material is joined to the beam, the piezoelectric film. Therefore, when a force is applied, the inertia of the seismic mass causes the beam and the piezoelectric film to bend in proportion to the acceleration. In the wearable device, the forces that cause the bending are the basic physical facts of walking I referred to earlier.

This bending turns into an electrical signal through a remarkable material property called piezoelectricity. For reasons that involve induced electrical dipoles, bending and deforming certain solid-state materials causes a change of electrical properties in proportion to the deformation. Though I do not know exactly what Fitbit™ or similar wearables on the market use, quartz (SiO₂), Berlinite (AlPO₄), and lithium niobite (LiNbO₃) are crystals that have been used in similar applications of piezoelectricity.

In abstract terms, this conversion of mechanical force to an electrical signal involves a transducer element. A transducer in electrical engineering converts one form of energy into another — in this case from mechanical to electrical via the deformation of the piezoelectric film.

Once we have an electrical signal we are not far away from digital data. A central component of a great many electronic devices is the analog-to-digital converter (ADC). The analog properties of the electrical signal, like its current and voltage, are converted into digits — the figurative ones and zeros — thanks to the inner workings of the ADC.

In practice, all of the above is packaged together inside a shell of plastic and referred to as an accelerometer. As with so much of electronics engineering, therefore, a specific functionality has been modularized, its physical details of the sensing and transduction covered up within a literal black box — a tiny one albeit, see for example this product flyer by Bosch™. The component is integrated into the wearable through standardized interfaces for data input and output and, if needed, an electrical power supply. Thanks to the modular design, accelerometers are mass-produced by specialized manufacturers and sold in big reels to big factories that assemble electronics products, often as contract manufacturers — an economic kind of modularity.

An archetypical electronics component with pins and a shell that covers the inner workings, such as the sensing of acceleration and the transformation thereof into digital data accessible through one of the pins.

Once the microprocessor of the wearable device receives the stream of digital data from the accelerometer, the signal processing, data analysis, and machine learning can commence, and the forces, their direction, magnitude, and time distribution are classified as steps taken.

This was a “teardown” of one available implementation of the accelerometer functionality down to the atoms, as it were. Many other components and functions constitute a complete wearable product and they too can be inspected down to the very atoms. A key practical point, however, is that I do not have to do so in order to design and construct a wearable step counter. The data scientist who generates the step data and a conditional push notification that prompts us cheerfully to “get up and move some more”, does not need to know about serial ports of accelerometers, even less about piezoelectricity and cantilever beams. Thanks to standard interfaces, the modularity in electronics engineering enables a great deal of abstraction and thus more efficient design, and, I suspect, more creative minds are empowered to build.

From a Biological Mixture to Test of Pregnancy via a Cascade of Antibody and Enzyme Interactions

The enzyme-linked immunosorbent assay (ELISA) has many variations and uses, making it an essential assay in the biologist’s toolbox. At its most basic it is a method that begins with a liquid mixture of molecules — e.g. blood plasma, urine, or the protein extract from a cell after cell lysis — then determines the concentration of a specific substance in the mixture. Applications include the detection of human chorionic gonadotrophin (hCG) in urine for pregnancy testing and the quantification of expression yields in different cell lines of a specific biologics cancer drug.

The basic physical facts in ELISA are the amount of the specific substance in the liquid mixture, or in other terms, the concentration of the type of molecule of interest that comprises the specific substance. This quantity is at the root of the utility of ELISA, which can yield answers to the questions: am I pregnant? or which cell line is preferred for mass production? As with the step counter, a transformative process is required to go from the former to the latter.

The typical 96-well ELISA microplate about to be placed in the spectrometer. The intensity of the color is indicative of the concentration of a specific molecule , such as hCG.

First, is the sensing element. In ELISA that is almost always an antibody, by convention referred to as the primary antibody. This antibody is engineered to bind strongly to one specific part of the molecule of interest, preferably with very high specificity and affinity, such that the antibody does not bind to other types of molecules in the mixture. Antibody engineering takes effort, and novel targets can be laborious to do well. However, for standard applications, like the detection of hCG, there are primary antibodies, like anti-hCG antibodies, already developed and validated and available for purchase from commercial suppliers like Thermo Fischer Scientific™ or Abcam™.

The sensing process in ELISA then proceeds with the primary antibodies binding to the molecule of interest if said molecule is present in the mixture, and in doing so they form a larger conjugate molecule. By some design that I will not elaborate on, the molecules of interest have already been “captured” before the addition of the primary antibodies and they are immobilized to a part of the lab equipment. The larger conjugate molecules are by extension also immobilized.

But sensing the molecule of interest with the primary antibody is not enough. The next step involves a transducer element. Different variants of ELISA use different methods. A common variant of ELISA involves the addition of a secondary antibody, which is engineered to bind to the primary antibody. Recall that following the sensing step the primary antibody is part of the immobilized conjugate molecule. Standard secondary antibodies are available for purchase.

The secondary antibody is equipped with one more property: it can produce a strong and specific optical signal. Traditionally this signal derives from the horseradish peroxidase (HRP) enzyme, which is covalently joined to the secondary antibody. The HRP of the secondary antibody catalyzes a reaction that modifies yet another molecule, the chromogenic substrate, which is added to the mixture. In its modified state, the chromogenic substrate generates a specific color. The organic molecule TMB is one example of a chromogenic substrate.

No shortage of moving molecules…

Many molecules are involved in ELISA. This cartoon drawing illustrates the sensing step (1.) followed by two transducer steps (2. and 3.), where the last one generates the optical output.

Ultimately the transducer element converts the number of immobilized conjugate molecules into a proportional optical intensity at a well-defined range of visible light. Scientists can measure that intensity with a spectrometer in the lab — or for simpler applications, just look at the mixture — and the answer to the question of pregnancy or cell line for mass production is obtained.

The measurement also implies the creation of data. As in the case of the step counter, the data can be uploaded to databases and become part of data analysis of trends or patterns. Laboratory robotics can be used to further automate the creation and storage of ELISA data.

I have described the ELISA process in terms that are more abstract than the perspective of a practitioner in the lab. Important practical steps of rinsing and diluting are left out. I have with this abstraction, however, mapped to an identical template of sensing and transduction, the transformative process of the step counter and ELISA, from basic facts to utility via the atoms. Biosensors are an engineering synthesis of said elements from these two archetypical designs.

The Biosensor Fountainhead: Glucose Sensing at Home

Diabetes is a major health problem — half a billion people worldwide are estimated to live with it, and demographics and diets mean this is likely to trend upwards. If the glucose level in the blood of the diabetic person deviates too much from baseline, preventive action must be taken promptly, such as self-administration of insulin. It is a disease that has to be managed. Nowadays, self-monitoring of the glucose level involves the collection of small blood samples, usually from the tip of the finger which the diabetic person pricks with a lancet needle. Under some conditions, this inconvenient procedure has to be done several times a day.

Leland Clark and Champ Lyons created the field of biosensors in 1962 when they invented the first portable sensor for glucose in the blood. The central component in their innovation was the enzyme glucose oxidase (GOx).

Though details in the implementation have changed since the 1960s, the enzyme-catalyzed oxidization of glucose is still the basis of the electrochemical reactions at the heart of many contemporary glucose sensors.

The transformative steps in an archetypical glucose sensor are:

  1. Glucose is oxidized to form gluconic acid. The oxidization takes place through the reduction of the enzyme cofactor flavin adenine dinucleotide (FAD), which is embedded within GOx. The reduction entails the transfer of two hydrogen atoms from glucose to FAD.
  2. FAD is reoxidized by another reaction. The nature of this reaction varies a great deal between sensors. One example is dissolved molecular oxygen (O₂) reacting with the two hydrogens from reduced FAD to form hydrogen peroxide (H₂O₂).
  3. Hydrogen peroxide is simple enough that it can be quantified with amperometry. The hydrogen peroxide is oxidized at a solid-state electrode, made of platinum for example, which leads to the release of electrons. Consequently, a flow of electrons is generated, that is a current, which is readily quantified with electronic components.
Cartoon representation of GOx, with the FAD cofactor highlighted (PDB structure 1GPE); protein diameter about ten nanometers.

Note how these steps involve both sensing and transduction. The glucose is sensed by the GOx enzyme through a highly specific protein-ligand binding event. The sensing element, therefore, exploits a naturally evolved enzyme and the biotechnology we have that enables the production and engineering of GOx. It is followed by a number of transformative steps that transduce the fact of a binding event until an electrical current is obtained at a standard solid-state electrode. So the more glucose is present in the blood, the more binding events take place, and the greater the measured electrical current.

In ways not that different from the step counter wearable outlined above, the magnitude of the electrical current is convertible into digital data. And thus our biosensor is complete.

Well… not so fast.

There are clearly many supporting elements that have to be in place for the figurative wheels of this biosensor to turn smoothly. To name one complexity: there are other molecules in the blood, some of which can interfere with the transduction steps. One known example is the painkiller acetaminophen (a.k.a. Tylenol™), which can interfere with the amperometry and amplify the electrical current. Therefore, another set of molecules or additional filtering may be preferred in the final construction of the glucose meter.

The practical complexities and their fixes aside, the glucose meter as outlined above was the first real-world biosensor. Product and technology development in this area has led to a great deal of commercial activity, where for example the acquisitions of MediSense™ and TheraSense™ by Abbott Laboratories™ for roughly $1 billion each, were in large part an acquisition of intellectual property related to this foundational biosensor.

The Non-Invasive Glucose Biosensor Moves Slowly Through the Product Pipelines

The active collection of blood is hardly ideal for diabetes self-management. Therefore product prototypes that use non-invasive and passive means to quantify glucose have been made. In fact more than prototypes. In 2001 the U.S. Food and Drug Administration (FDA) approved GlucoWatch™ by Cygnus™, which monitored glucose levels in interstitial fluid extracted automatically through the skin but without breaking the skin. However, the GlucoWatch™ failed to gain traction (skin irritation, excessive alarms, and lack of reimbursement are some reasons given) and neither Cygnus™ nor GlucoWatch™ exists anymore.

The first FDA-approved non-invasive glucose biosensor. The product is no longer sold because of problems.

Sweat has in principle many benefits as the analyte for the biosensor. Sweat is always available and in large volume, relatively speaking. Our skin emits sweat to regulate heat. Though we are mostly aware of our own sweating during exercise and on a hot day under the sun, there is sweat on our skin at other times as well. Ions and small organic molecules — like glucose and ethanol — exist in sweat, and in concentrations that relate to the concentrations in blood. So if glucose can be reliably sensed and quantified in sweat, we have the first element of a biosensor for diabetes management without the bothersome finger pricking.

A detailed “works-like” prototype of a sweat-based glucose biosensor, all the way to the smartphone app front-end, was disclosed in 2016 by a team at the University of California, Berkeley. Glucose oxidase is again at the heart of the sensing element, as are a similar set of reduction and oxidization reactions that ultimately lead to an electrical current. However, sweat, not blood, is the fluid that is in contact with the enzyme.

The electrode in this prototype is a thin piece of gold, 3 millimeters in diameter, to which the enzyme has been immobilized. The electrical current that arises at the electrodes is turned into digital data by a microcontroller, which in turn communicates the data over a standard Bluetooth connection to a nearby smartphone. The electrode, microcontroller, and Bluetooth antenna are all integrated into a flexible printed circuit board (FPCB), which can be wrapped around the wrist.

The sweat-based prototype of wearable glucose biosensor. From patent application, US20180263539A1 (abandoned).

The corresponding patent, however, has since been abandoned as there were a great many prior inventions disclosed for the various parts of the prototype. This suggests that most if not all of the primary functional parts for a device like this are around — what is missing is the synthesis of the parts into a commercially viable, consumer-grade product.

Another non-invasive approach that has been explored uses the ocular fluid, or tear fluid, which lubricates and nourishes our eyes. Like sweat, it is always present. Verily Life Sciences™, part of Alphabet™, and Novartis™ partnered in 2014 to develop a contact lens that could monitor glucose in the tear fluid of humans. In 2018, however, the effort was put on hold because of interference caused by other biomolecules in tears.

I do not know the exact construction that was tested by Verily™ and Novartis™, but related patents from Verily™ mainly contemplate the sensing element to be the glucose oxidase biosensor as described above. Many additional engineering challenges had to be overcome in order to integrate the biosensor into a contact lens, including the ability to transmit data from the lens (picture a Bluetooth antenna in your eye). Still, it was not enough, at least not up until 2022.

These are far from the only prototypes of non-invasive glucose self-monitoring. They are illustrative though of how biosensors can be used and integrated with electronics. Despite the engineering advances, non-invasive glucose monitoring in a mobile or wearable format has not yet been successfully productized.

Biology Sensory Superpowers and Antibody Engineering Meets Nano-Electronics Engineering For Transduction

If biology was a superhero with superpowers, our hero could do specificity like no other. Either by natural selection or by human design, a macromolecule can almost always be created such that it binds to another macromolecule and only that macromolecule. That is the specificity of binding.

Macromolecules, like proteins, DNA and RNA, are comprised of a variable number of physically diverse residues, like the twenty amino acids and the five nucleotide bases. As a consequence, macromolecules can form a vast number of distinct surfaces thanks to the shape and electrostatic variability of the amino acids and nucleotide bases.

As shown above, in ELISA and the glucose meter, a specific binding event between molecules can be the molecular foundation of sensing. The specificity superpower is one reason why antibodies have become the workhorse in biologics drug development. Many tools of protein engineering have also been developed to create antibodies that can perform highly targeted action within the human body. In oncology, that function is the specific and efficient destruction of cancer cells — a major part of the bioeconomy I have written about before. These engineering tools can easily be repurposed for the creation of biosensors.

But a very specific sensing element is never enough — by itself it is a one-sided superpower. Transduction is required as well. In ELISA, as we saw above, the transduction involves a range of secondary molecules and ultimately generates an optical output signal. An electrical signal is however preferred when integration with electronics is part of the product specification.

One published biosensor approach builds on a design that couples an antibody binding event directly to an electrical signal. At the core of this approach, there is textbook electrochemistry with electrodes and a flow of charge that follows from redox reactions. It has that in common with the glucose sensor I described above, but it differs in how the binding event alters the electrochemistry.

Two electrodes and an analyte fluid between. The electrode on the left is carbon-based to which primary antibodies (green) have been immobilized. The only major molecular change near the carbon-based electrode takes place if the specific molecule (purple) is captured by the primary antibody.

As in ELISA, there is a molecule of interest. In this specific study, it is cortisol, but it can in principle be something else. An off-the-shelf antibody that binds specifically to cortisol is purchased. I call this the primary antibody in order to make the parallels to ELISA clear. The primary antibodies are immobilized to a carbon-based electrode. Functionally speaking, this electrode is strong, yet flexible, just like a synthetic polymer in clothing, and it has been prepared, such that it can conduct electricity and participate in redox reactions. The electrode is also coated with a blocker molecule, like the albumin protein, which ensures that no (or at least very few) molecules in the analyte are immobilized to the electrode unless there is specific binding to the primary antibody.

The flow of charge in this design can be obstructed to a degree that varies with how much molecular matter is immobilized around the carbon-based electrode. Simply put, all that organic molecular matter is a hurdle for the charge flow to overcome and thus constitutes a form of insulation. The greater the number of conjugate molecules between the primary antibody and the molecule of interest, the greater the change in the electrical properties of the electrochemical cell. With some standard voltammetry, the altered dynamics of the electrical current between the electrodes are quantified and related to the number of bound molecules of interest.

There are many other components and designs of appropriate cost, energy consumption, and footprint that have to be in place in order to turn this molecular process into the transducer element of a commercially viable wearable biosensor. This is nonetheless a process that both exploits the specificity of antibodies and directly generates electrical signals. The flexible yet sturdy nature of the carbon-based electrode is attractive when we brainstorm designs of flexible wearables, perhaps ones part of our “smart” clothing even.

Messy Molecules, Modularity, and the Economics of Wearable Biosensor Product Development

Since at least the days of Achilles, a hero’s strengths come at the cost of other weaknesses. So too for biosensors.

The basic science of glucose sensing, established by Clark and Lyons in 1962, has not yet been adequately transformed into a non-invasive wearable consumer product with a product-market fit — though, I think, optimism for the near future is warranted for this specific case. The electronics sensors, on the other hand, have from the 1960s to the present gone from bulky and energy inefficient to a mass-produced commodity that can be packaged inside small devices powered by coin cell lithium batteries, which engineers and designers instruct through software abstractions and standard communication protocols.

One inherent challenge of biosensors is biofouling. Biological matter has a remarkable ability to stick to surfaces and clog up spaces. The wet and heterogenous nature of macromolecules creates thermodynamic conditions that favour gradual aggregation. As anyone who has stored a protein solution knows, degradation is inevitable, and tricks and additives are available to attenuate the effect. But entropy is relentless and biology is a battle against decay.

Gradual decay is certainly not unique to biology. The rechargeable lithium battery for example becomes less rechargeable with use and time because atoms in the battery can move in more ways than the one that affords the battery with its energy storage utility, and the ubiquitous semiconductor as well decays with time in part due to that the nanoscale structure drifts over time.

The difference is rather with respect to time scale. If the energy barriers to competing processes are lower, as they often are in a condensed phase, biosensors will have more restrictive expiry dates, so to speak. Then as we dip the biosensor into blood, sweat, and tears — the real stuff, not just the figurative — the number of competing processes that involve other molecules becomes sizeable.

Another challenge — and opportunity — is modularity. Electronics design and engineering are helped a great deal by that components and parts are sufficiently general such that, for example, accelerometers in wearable step counters, cars, and virtual reality goggles, can be of an identical design or at least use an identical communication interface. At the component level in electronics not much has to be special purpose design. The differentiation between most electronics products rather is in the product design, application, and quality of the whole of the numerous utilities.

Economic incentives have helped make it that way. Could biosensors follow the same path once standards and best practices emerge from a plethora of prototypes, special purpose designs, and bioelectronics hacking? It might. But the fact that the superpower of biosensors derives from finely tuned protein and nucleotide sequences, all the way down to the atoms, means any given structure is far less likely to have transferable utility. In what sense, and at what marginal cost, could a well-designed glucose biosensor be reconfigurable to enable the sensing of cortisol, human chorionic gonadotrophin, cocaine, etc.? Or, how much of a well-designed glucose biosensor could be part of a different biosensor, such that the economies of scale for one aids the other as well?

In order to answer these questions in ways that move bioelectronics product development closer to solid-state electronics product development, innovation no doubt is needed. Could a near-universal transducer element be possible, for example? There is also the market pull to consider. The home desktop computer, then the laptop and smartphone generated major incentives to supply the electronics product developers with cheap, good, and general components. Is diabetes self-monitoring that major market pull, which can provide the positive externalities that benefit all consumer bioelectronics utilities? Or is there a premier biosensor application waiting to be discovered — the killer app, so to speak?

I have described only a few specific designs of biosensors over the course of this piece. There are many more designs out there, especially academic prototypes. A great deal more remains to be done though, especially with respect to how to scale prototypes to large, global consumer markets. The cross-fertilized consumer product I imagined in the first paragraph has not yet come about. However, the atomic foundations of its many parts are beginning to emerge. To unite them all into a product design with market fit is itself an art and opportunity.

Anders Öhrn is a builder of antibodies and electronics and the tools to build them. First, in a Vancouver biotech startup, he developed and applied computational tools to create cancer drug biologics. Second, in a Toronto consumer electronics startup, he developed smart home consumer electronics. The foundation is intellectual curiosity, business drive, and the belief that the most useful things are created at the junctions of conventional domains. He also develops analytical tools, machine learning included, to make sense of the great volume of data any effort nowadays produces. His search for new ventures is as active as ever.

Follow Bioeconomy.XYZ, in order to learn more about all the ways biotech is shaping the world around us.

Join Bioeconomy.XYZ’s new Slack Community to build better connections in the new hybrid world and to foster more discussion around the most pressing topics our writers and readers are surfacing.



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Anders Ohrn

Quantitative if possible, towards first principles, pragmatic always. Innovation, biology, computation & complexity.