The first design was conceived around a number of theoretical guesses that we needed to prove experimentally.
Our first prototype was born to test these assumptions. Some of them worked as expected, but many of them only served as an initial step towards a more concise device.
In the following paragraphs, we want to share our experience working with a new device, an initial prototype, designed by us, and tested to the limit. By learning how to set our prototype up, we were learning about every factor involved in the process.
Designing, manufacturing, and testing our devices have been great overall. Engineering biodevices require theoretical and experimental skills. But what we have learned is the most simple lesson: the best way of doing anything is doing while learning.
Our first assumptions
Binding aptamers to a PDMS surface. In order to create an electrostatic and mechanical trap for our target protein, we planned to work in a PDMS environment. PDMS is a well-known manufacturing material for electronics. So we could easily integrate PDMS in our device.
Optical measurement sensor. The materials required to test our sensor were a 280nm UV LED emitter and an LDR. The amount of light traversing the solution was quantified by a drop in voltage across the LDR: with higher protein concentrations, higher absorption is expected together with an increased drop in voltage.
Microfluidics: for channelling fluids through the chip. Microfluidics allows us to move samples on a microliter scale, minimizing the dead volumes and the waste through the chip.
Modular design and normalization: We needed to standardize the protocols related to hardware to reduce the number of variables involved. This would restrict the design and manufacturing process and help us a lot when playing with certain design parameters.
Enable the DIY: We had the need of developing everything in a way such that anyone, regardless of their origin, could replicate our experiments in an affordable and creative way.
How it works
The device is divided into four parts. Every subsystem has been conceived to be integrated into a bigger whole: the device.
Custom modules oriented to experimentation.
Custom PCB created specifically for our device.
Microfluidics and aptasensor
Affordable PDMS chip manufacturing.
Input/output and chambers of measurement normalised for a chip design oriented to manufacturing.
Immobilised aptamers on PDMS surface.
System of measurement
Optic system of measurement based on protein absorbance at 280 nm of wavelength.
LDR as light receiver.
Pressure pump regulated manually
From the point of view of the protein sample, the fluid would follow this stream:
The protein sample starts inside the syringe, suspended in a buffer solution. And it is manually pumped into the PDMS chip.
From the inlet, it is displaced to the sensor chamber, where the surface is almost covered by aptamers.
The circulating proteins get trapped in the aptamers chamber.
A clean buffer solution is injected into the chip, to wipe away any other molecule.
A saline solution is pumped into the chip, to detach the protein from the aptamer. The process has been previously described in this this paper.
The saline solution with the remaining protein is taken to another chamber, where absorbance is measured at the 280 nm wavelength.
Once the measurement is made, we get the information and analyze it.
The chip is cleaned and the process can start over again.
The cassette is a two-part PMMA structure, able to allow 4 fixed inputs (1 mm of diameter) and 4 fixed outputs (1 mm of diameter), 4 chambers for 280 nm UV LED modules and 4 chambers for LDR modules. An open module was designed for inserting other components if required by the user. Find more information on the modules in our our GitHub.
The control electronics were designed to govern 4 LED modules and 4 LDR modules.
It has a frame for the 40x40 mm PDMS chip, which relates to the bed of the laser cutter. PDMS chip should be 3 mm thick.
These parts are fixed together with screws and spacers. Although not a large pressure is required.
The chip design
The chip has dimensional restrictions (40 x 40 mm) due to the boundary condition of manufacturing: the laser cutter bed maximum dimensions.
We created polymerization chambers for this purpose.
The input and output have been fixed for manufacturing concerns.
The last chip that we designed has two parallel circuits. Therefore, two inputs and two outputs and four chambers (two per circuit).
The chambers were conceived to be in between an emitter and a receiver, facing one another.
The electronics for an absorbance related measurement
The electronics are governed by an Arduino Nano board. It connects the analogue electronic board inside the device and the Arduino IDE. So the user can see the data related to the signal in the serial monitor.
The light source is controlled by the Arduino through a 2N2222 transistor for providing the module with 6.5V and enough current.
The circuit which receives the data related to the signal comes from the light detector. When the protein sample contains a high concentration of protein, the voltage drops proportionally to the amount of protein. Then, the signal is amplified and corrected via an OpAmp and an Instrumental Amplifier.
The voltage is measured by the built-in Arduino Analogue to digital converter.
Further information can be find in our GitHub.
Measurement results: Optical Detection of Protein
An initial approach to optical detection of OLE1 was to be by means of UV absorption at 280nm, similar to how a spectrometer quantifies the amount of protein.
An inexpensive 280nm LED was bought, together with a light dependent resistors (LDR). The first prototype was built such that the LED would shine across a cuvette containing a solution with unknown quantities of protein. The amount of light traversing the solution was quantified by a drop in voltage across the LDR: with higher protein concentrations, higher absorption is expected together with an increased drop in voltage.
Analog Signal Processing
The first tests were carried out with an LDR is placed in front of the 280nm UV LED. The drop in voltage was measured by means of an instrumentation amplifier and a low pass filtering circuit (2 Hz cutoff) as shown in the image below. The values of the amplification had to be tuned according to the measurements being recorded.
The first tests with known highly-concentrated protein solutions of BSA (13.67mg/mL) and binding buffer yielded the following results:
The results above show that the points can be distinguished (the error bars do not overlap) for high concentration changes of 10 orders of magnitude. Purifying that amount of protein with aptamers would be unviable. To test the limit of detection further and with lower concentrations of protein, the 0.1367 mg/mL solution was used for taking readings after diluting it again in the same proportions.
The plot above shows that when BSA was diluted to concentrations of 13.67, 1.367 and 0.1367µg/mL or less, changes in luminosity (UV absorption) were indistinguishable (non-significant). Note that the gain has been readjusted with respect to the previous readings.
Time dependency was evaluated by taking long recordings under stationary conditions to assess the system response. In an ideal case, the mean value would remain static surrounded by some stochastic errors and interferences. When measurements of around 10 seconds are taken, there is no appreciation in signal decay as seen in longer recordings, but recording of 5 or 10 minutes revealed a time response that had to be accounted for (see figure below).
To try and improve the limit of optical detection without the need of more expensive and higher quality equipment, digital post-processing was built into the microcontroller to try and increase the quality of the signal.
We had seen that very short measurements (10 seconds) and also extended measurements (minutes) were subject to higher errors. Short measurements were very sensitive to stochastic and high frequency noise, while longer reading experienced a decay in the signal coming from the hardware (either from the LED or the LDR).
The first stage of digital processing, an averaging filter, produced positive results (see figure below). Digital processing with a running average filter smoothens the values obtained and narrows the error bars. Note that the analogue circuit has been adjusted to maximize the amplification within the range of interest but normally the time decay curve would not be as noticeable.
Characterizing the system’s time decay response
To characterize time decay, several 5 minute measurements were taken (after switching off the circuit for some time to leave it to rest). The filtered data points measured were very consistent across all measurements showing the same decay behavior. The data points are shown in blue in the plot below together with a fitting curve defined by the equation L(t) in black.
With the decay curve characterized from the empirical data obtained, this information can be used to correct this artifact. One way of doing it would be to consider the signal recorded, Srecorded, as the product of the decay curve, L, and the non-attenuated signal, Sstable, and solve by finding the inverse function, L-1.
However, given that inverse functions can have no solution or may cause stability problems (for example, a denominator that goes to zero), the most viable and simple alternative is to express time decay as a summation component instead of a multiplication one.
This equation can now be solved to correct for the time decay artifact (see figure below).
Since all signals are subject to the same time decay effect, a reference LDR can be used to correct for the effects of this artifact. Each LDR needs to have the same analogue circuit, and the signals are then combined during post-processing. Since measurements do not have to be real-time, post-processing is not a big issue if simple computation algorithms are used.
Note that the reference LDR and the signal LDR do not record exactly the same since they do not receive the same amount of light: one is physically placed at a different opening in the chip, in front of a different LED. However, the gains can be adjusted so that both show the same time decay curve. The information from the reference LDR - which remains under stationary conditions - can be used to correct the signal LDR while measurements of protein concentrations are taken.
However, even though the magnitude of the variance remains constant, the variance increases with respect to the mean with time. This increase in the coefficient of variation is not good because the relative error increases with time.
Testing the device
To test the final device, sequential flow measurements of around 10 seconds were taken in this chip with two LEDs and two LDRs and high concentrations of protein and a pipette. Ideally, these would have been cycled with the syringe pumps, but the control software was not fully ready yet.
After analog and digital processing, we were unable to improve the limit of detection beyond the values shown in the initial figures: 13.67µg/mL of BSA and changes of 10 orders of magnitude in the concentration. A better limit of detection of UV absorption would be required to measure airborne allergens captured by aptamers. The quality of our equipment should be improved but at a prohibitive cost, at which point we decided to center our effort to potentiostat detection where the literature suggested finer measurements were achievable with a relative inexpensive devices.
Things we learned while doing
The most important lesson we obtained was about the system of measurement. As we had no relevant results in this part, we proceeded to change our approach, due to the following reasons:
The initial PDMS chips we made were very unstable. Troubleshooting leakages and integrating feasible input/output fittings require patience and creativity. We learnt the following:
Modular design made our lives much easier. Designing a chip, cutting it with the laser, curing the PDMS in the polymerization chamber and integrating the fittings was very easy, as we had developed an standard way of doing it.
It was not worth it attaching aptamers to the PDMS surface, as the range of detection of the sensor was too far from tracing our concentrations.
DIY was the way to go, as we could not spend time and money buying commercial equipment and learning how to use it.
- The sensor, at 280 nm, did not have a detection limit high enough for our necessities. It could not trace our concentrations.
- Optical-based sensors are too sensitive to ambient conditions. So we should refuse to use it in our final design.
- Optical-based sensors are too sensitive to metrologic precision parameters, as emitter-receiver alignment for instance.
- Our system should integrate a straightforward way of experimenting with microfluidics.
- We needed a smooth way of pumping microvolumes into the microfluidic system. PDMS is too sensitive to mechanical parameters, as pressure or input/output torques.