A Brief Guide into Single-cell Analysis

Intro: Seeing the Single Cell?

The evolution of analytics in biology has reached a new level with single-cell technologies. Unlike bulk sequencing, these platforms let us look at biology one cell at a time, how it’s behaving, what it’s expressing, and where it sits in a tissue. It’s a widely used method, one I’ve come across in awe multiple times in my scientific career. But the way it actually works? That’s what always fascinated me.

In this post, I wanted to take a step back and understand how exactly these machines work on a physical level, and to do so in a digestible way. I’m certainly not claiming expert status, just sharing the perspective of a relentlessly curious engineer!

Case Study: Mapping the Tumor Microenvironment with CODEX

I’ve found that understanding complex technology becomes much easier when placed in a real-world context. That’s why I want to frame this technical dive through a case study of CODEX (CO-Detection by indEXing), a platform that’s not only visually stunning, but also just a marvel in optics and engineering.

I’m biased, of course. My appreciation for CODEX comes largely from the work of **John Hickey**, whose papers and datasets on tumor–immune interactions show just how much insight you can extract from a single tissue section [1].

Breakdown of Protocol

The process starts with tissue slides containing a sliver of solid tumor tissue embedded in parafilm. Like in traditional immunofluorescence, CODEX uses a cocktail of up to 40–50 DNA-barcoded antibodies, each one targeting a specific protein on or inside the cell. But here’s the trick: the antibodies themselves don’t fluoresce. Instead, they’re tagged with unique DNA barcodes that can be read out in cycles using complementary fluorescent probes.

Once the tissue is stained and mounted, the CODEX system kicks in to extract high-dimensional data from a static piece of tissue. Unlike traditional microscopy, CODEX doesn’t try to image all markers at once. Instead, CODEX uses cyclical imaging: in each cycle, only 2–3 markers are visualized. These markers are revealed by adding fluorescent oligo probes that temporarily bind to DNA barcodes on the antibodies already stuck to proteins in the tissue.

Each cycle looks like this:

• Add probes (fluorescent tags)
• Image with lasers
• Strip the probes
• Repeat with a new set of probes

Iterative Imaging Workflow in CODEX [2]

CODEX Optics

What really intrigued me was understanding how the microscope physically captures an image. During imaging, the tissue is illuminated by precisely controlled solid-state lasers. Each laser is tuned to a specific wavelength that matches the excitation spectrum of a fluorophore, essentially lighting up only the molecules we’re interested (commonly DAPI, FITC, Cy3, Cy5 — so ~405, 488, 561, and 647 nm lasers).

Each pixel in the image is created by exciting these fluorophores at precise spots in the tissue and measuring the light emitted as they relax. That light is collected through:

• A high NA objective lens (usually 20x or 40x water or oil immersion)
• A dichroic mirror that splits excitation from emission
• Emission filters that isolate specific fluorescence bands
• And finally, a sCMOS or CCD camera that records the signal intensity

This is done across every field of view, sometimes hundreds per tissue, and often across multiple z-stacks (depths) for 3D reconstruction.

Simple breakdown of the CODEX laser technique

CODEX Fluidics

The other main mechanic lies in the fluidics system, which allows the machine to go through the wash cycles mentioned earlier. The job of the automated fluidics system is to control the incubation on the tissue and to remove the fluorescent probes cleanly. The delivery of probes is governed by laminar flow, described by the Navier-Stokes equations, where fluid moves in smooth, predictable layers across the surface of the slide as a way to “wash” the slide of the antibodies. This ensures reagents spread evenly without turbulence, which could lead to uneven staining or physical disruption of the tissue.

Then, with the application of Fick’s law, the machine allows time for the hybridization process based on the dispersion time of molecules spreading over the tissue. This timing is important, too short, and binding is incomplete; too long, and background fluorescence can rise due to nonspecific interactions or photobleaching under ambient exposure. Each cycle is a balance of shear stress and surface tension, as the flow must be strong enough to remove unbound fluorophores, but gentle enough to avoid dislodging cells or washing away bound complexes.

CODEX Fluidics System for Tissue Washing [2]

The final step touches on the computational aspect of the system, in which the images are stitched together and aligned based on nuclear landmarks seen with the DAPI staining. The machine also uses autofocus algorithms to maintain sharp imaging across cycles and z-planes, correcting for slight shifts or warping over time. Otherwise, even minor drifts in focus can distort downstream quantification. Speaking of, this level of machinery is handling terabytes of grayscale TIFF stacks, one for each marker and z-plane across each field of view, that are sorted and analyzed using computational methods to make the data truly single-cell…which deserves a post of its own.

Final Remarks

Single-cell technologies are incredibly powerful tools with growing influence in medicine. In cancer, a field of my interest, these tools are being used to track clonal evolution, map spatial architectures, and identify rare resistant subpopulations that drive relapse. They’re also key to optimizing next-generation cell therapies, like CAR-T and TIL-based approaches, by enabling precise functional and phenotypic screening of these cell products.

I wrote this piece to highlight just how much innovation goes into the creation of scientific data, especially in the machinery that data is generated from. In a field that’s quickly drawing attention from people outside its usual circles, it’s easy to overlook the behind-the-scenes complexity. And when that complexity gets ignored, the science itself can sometimes be met with skepticism. My hope is that my curiosity can shed light on the technical beauty of these systems and be a way to educate those who might be new to these frontiers of science!

References

[1] https://doi.org/10.1038/s41596-021-00556-8

[2] https://www.leinco.com/codex-technology/