Iris sat at the laboratory bench. Iris looked at the digital screen. The digital screen showed a graph. The graph contained six lines. The six lines represented six different runs of the same sample.
The six lines did not match. The first line was high. The second line was lower. The sixth line was in a different place entirely. Iris felt a cold sensation. Iris had spent on this experiment. The sample was stable. The reagents were new. The laser was turned on. The laser was warmed up.
14%
Total Drift
Visualization of the shifting data points between and .
Iris had calibrated the system at . It was now . The data showed a drift. The drift was 14%.
A Crisis of Reproducibility?
Iris opened a file on the computer. The file was a PDF. The PDF was about the reproducibility crisis in biology. The PDF said that 70% of researchers fail to reproduce experiments.
70% Research Failure Rate
Standard industry metric for experimental non-reproducibility.
The PDF blamed poor statistics. The PDF blamed contaminated cells. The PDF blamed the pressure to publish. Iris read the PDF. Iris wondered if she was a bad scientist.
She checked her pipetting. She checked the buffer pH. The buffer pH was 7.4. She checked the room temperature. The room temperature was . Everything was correct. Iris looked at the flow cytometer. The flow cytometer was a large box. The box was silent.
Iris decided to clean the system again. Iris took a bottle of cleaning solution. Iris ran the cleaning solution through the tubes. Iris ran distilled water through the tubes. Iris ran the sample again.
The seventh line appeared on the screen. The seventh line was lower than the sixth line. The drift continued. Iris reached for the manual. The manual was thick. The manual did not explain the drift.
Iris thought about the moon. She thought about the phase of the moon. People said the moon affected experiments. Iris did not believe in the moon theory. Iris believed in physics.
The Acoustic Engineer’s Entry
Iris called Ana Z. Ana Z. is an acoustic engineer. Ana Z. understands how materials vibrate. Ana Z. understands how materials join together. Iris told Ana Z. about the drift. Iris told Ana Z. about the flow cell.
Ana Z. came to the laboratory. Ana Z. looked at the flow cell. The flow cell is a small piece of glass. The flow cell has a channel. The channel is where the liquid flows. The laser hits the liquid in the channel.
“The flow cell must be clear. The flow cell must be precise. How do you clean this cell, Iris?”
– Ana Z., Acoustic Engineer
Iris said she used a standard detergent. Iris said she used a sonic bath. Iris said she cleaned the flow cell every day.
Why the Bond Fails
Ana Z. explained the bonding process. This is how the process works. A flow cell is made of separate pieces of glass. The pieces must be joined. Some manufacturers use adhesive. The adhesive is a glue.
Adhesive Layer
Reacts to chemicals & vibrations. Expands unevenly.
Pure Glass
Inert, stable, and uniform expansion.
The glue holds the glass together. The glue creates a seal. But the glue is a different material than the glass. The glue has a different thermal expansion. The glue reacts to chemicals.
When Iris uses the detergent, the detergent touches the glue. When Iris uses the sonic bath, the vibrations hit the glue. The glue begins to fail. The failure is small. The failure is microscopic.
The glass does not fall apart. But the channel changes shape. The channel expands by a few microns. The light path changes. The light path is no longer consistent. This causes the drift.
Iris looked at the flow cell. The flow cell looked perfect. Iris could not see the failure. Iris used a microscope. Even under the microscope, the glass looked clear. But the data said the glass was failing. The data was the only proof.
Iris realized she had been intellectualizing a hardware defect. Iris had thought about the reproducibility crisis. Iris had thought about complex biological interactions. Iris had thought about software errors. Iris had looked for a profound explanation.
The explanation was not profound. The explanation was a failing bond in a piece of glass. The bond was weak. The bond could not survive the cleaning cycle.
Foundations of Precision
Iris searched for a better component. Iris found
Iris read about their manufacturing. HookeLab makes high-precision optical components. HookeLab uses different bonding technologies.
Adhesive Bonding
Standard glue-based assembly.
Powder Fusion
High-heat structural bonding.
Optical Contact
Molecular-level fusion.
Optical contact bonding is a specific process. The process requires two glass surfaces. The glass surfaces must be polished to a high degree of flatness. The flatness must be within a fraction of a wavelength of light.
The surfaces must be cleaned of all dust. The surfaces are brought together in a clean environment. When the surfaces touch, molecular forces pull the glass together. The two pieces of glass become one piece of glass.
There is no glue. There is no adhesive layer. The bond is as strong as the glass itself. The bond does not react to detergents. The bond does not fail in a sonic bath. The thermal expansion is uniform because the material is uniform.
Iris ordered a new sheath flow cell from HookeLab. The new flow cell arrived in a small box. Iris installed the new flow cell. Iris aligned the laser. Iris prepared the same sample.
Iris ran the experiment. The first line appeared on the screen. Iris waited one hour. Iris ran the experiment again. The second line appeared. The second line was exactly on top of the first line.
After and three consecutive runs, the new hardware achieved absolute baseline stability.
Iris waited . Iris ran the experiment a third time. The third line matched the first two lines. The drift was gone. The drift was 0%.
Beyond the Philosophy
Iris felt a sense of relief. Iris also felt a sense of frustration. Iris had wasted . Iris had blamed her own skills. Iris had blamed the state of science. Iris had read long papers about the philosophy of reproducibility.
All those papers were irrelevant. The problem was a mundane hardware fact. The problem was a cell whose properties shifted with handling. The hardware was hiding inside a problem that everyone insisted was profound and abstract.
“People focus on the liquid. People focus on the light. People forget the glass that holds the liquid. If the glass is not stable, the data is not stable. The glass is part of the measurement.”
– Ana Z.
Iris looked at the old flow cell. The old flow cell was in the trash. The old flow cell was an example of a common mistake. Scientists prefer grand explanations. Grand explanations flatter the difficulty of the work.
If the experiment fails because of a deep biological mystery, the scientist is still an explorer. If the experiment fails because of a cheap glue bond, the scientist is just a person with a broken tool. It is hard to admit the tool is broken when the tool looks fine.
The lab manager came by the bench. The lab manager asked about the reproducibility crisis. The lab manager asked if Iris needed new software. Iris said no. Iris said she needed better glass.
Iris explained the bonding. Iris explained that the cleaning cycles were destroying the old cells. The lab manager looked at the cost. The HookeLab flow cell cost more than the old flow cell.
But the old flow cell cost of Iris’s time. The old flow cell cost the reagents. The old cell cost the trust in the data. The price of the better flow cell was low compared to the price of bad data.
Journal Note: Hardware Integrity
“Hardware integrity is the foundation of reproducibility. Inspect the hardware before questioning the biology. A measurement is only as good as the glass it passes through.”
Iris wrote a note in her lab journal. Iris wrote that hardware integrity is the foundation of reproducibility. Iris wrote that one should inspect the hardware before questioning the biology. Iris felt like a better scientist now.
She was not a victim of a grand crisis. She was a person who understood her tools. She understood that a measurement is only as good as the glass it passes through.
The room remained at . The buffer pH remained at 7.4. The laser stayed at 488 nanometers. The data stayed consistent. Iris finished her work. Iris went home. Iris did not think about the moon.
Iris continued to use the new equipment for the rest of the month. The results stayed the same. Each day, the lines on the graph overlapped. The overlap was perfect. Iris shared her findings with the department.
Other researchers were having the same trouble. One researcher was studying blood cells. One researcher was studying plankton. Both researchers saw drift in their data. Both researchers had blamed their samples. Both researchers had blamed the software.
Iris showed them the flow cell. Iris explained the optical contact bonding.
Institutional Shift
The department decided to change suppliers. They looked for components that could survive the lab routine. They looked for components with tight tolerances.
They found that many large suppliers did not want to make custom parts. Many large suppliers only sold standard catalogs. But HookeLab accepted the small orders. HookeLab made the specific geometries needed for the older machines in the basement.
This allowed the department to fix old instruments instead of buying new ones.
Finding the Truth in the Glass
Iris realized that the reproducibility crisis is often a collection of small technical failures. When a thousand labs have a thousand small hardware failures, it looks like a single large cultural problem.
It looks like a crisis of logic. But it is often a crisis of manufacturing. It is a crisis of glue. It is a crisis of surfaces that are not flat. It is a crisis of components that cannot handle a detergent.
Iris sat at her desk. She closed the PDF about the reproducibility crisis. She did not need to read it anymore. She had found the answer in the glass. The glass was no longer lying. The glass was telling the truth.
The experiment was a success. Iris turned off the laser. Iris turned off the computer. Iris left the lab. The lab was quiet. The hardware was stable.