From Microscopes to Models: My Research Journey with OPAls
During my time with OPAls, I had the privilege of working on a project that combined biology, image analysis, and automation. Our primary focus was on analyzing rat kangaroo cells to identify UFBs (Ultra-Fine Bridges) — tiny DNA threads that may have major implications in cancer biology.
What began as microscopy work evolved into an engineering problem — one that taught me resilience, adaptability, and the power of iteration.
🧬 Understanding UFBs
UFBs (Ultra-Fine Bridges) are thin DNA structures that appear during cell division (particularly in anaphase). They are not coated with histones, making them difficult to visualize and distinguish from other structures like tethers.
Why do they matter?
- UFBs are often found in stressed cells, which mimic the environment of cancerous cells.
- They may be biomarkers of genomic instability, and studying them could deepen our understanding of how cancer develops.
Our goal was to provide strong, consistent evidence that differentiates UFBs from tethers using image analysis.
🧪 Experimental Setup
We worked with rat kangaroo (PtK2) cells, which have large, easily observable chromosomes.
Stressors used:
Aphidicolin: Induces replication stress by inhibiting DNA polymerase.Colcemid: Disrupts microtubules, interfering with mitosis.
Stains applied:
PICH: Labels UFBs.DAPI: Stains all DNA in the nucleus.
After stress exposure, cells were fixed and stained, then imaged using high-resolution microscopy. We collected over 800 images across multiple conditions.
🧠 The Challenge: Manual Image Analysis
Each image had to be analyzed through multiple steps:
- Identify the cell phase, focusing only on anaphase.
- Apply thresholding to isolate stained regions.
- Manually select cells and bridges.
- Measure features like bridge length, intensity, and cell shape.
But we faced several problems:
- Some images were blurry or hard to interpret.
- Incorrect phase identification led to wasted analysis.
- The volume of data was overwhelming.
- Manual errors introduced inconsistencies and slowed progress.
⚙️ My Solution: Automation
I realized this manual workflow wasn’t scalable, so I decided to automate the process. I developed a custom script that handled the entire pipeline.
🚀 Features of the Automation System
- Phase filtering: Detected and focused on anaphase cells using morphological features.
- Thresholding and watershed segmentation: Isolated cells and bridges.
- Automated selection: Identified UFB-like structures based on geometry and intensity.
- Measurement extraction: Collected area, intensity, length, and position data.
- Batch processing: Ran across hundreds of images efficiently.
- Visualization tools: Output charts and tables summarizing results.
What started as a semi-automated system with human validation at each step became a fully autonomous pipeline after several iterations.
🔁 Iteration with Feedback
Throughout the process, I worked closely with my advisor, Dr. Veronica Gomez, who guided improvements:
- She emphasized batch processing, which led me to optimize image I/O and memory usage.
- We iterated on cell segmentation quality, reducing edge-case errors.
- Her feedback helped me make the tool more robust and user-friendly.
This back-and-forth taught me the value of scientific communication, even when the solution is technical.
🧰 What I Learned
This project helped me develop and strengthen several skills:
- Problem Solving: I recognized a bottleneck and took initiative to fix it.
- Scientific Communication: I explained technical decisions and data clearly to my biology-focused team.
- Software Engineering: I built, tested, and iterated on a real-world automation tool.
- Resilience: I faced errors, false starts, and late nights, but kept improving the system.
Most importantly, I learned how biology and computation can come together to enable deeper insights.
📈 Potential Directions
Although our project concluded with strong results, here were some potential extensions:
- Use machine learning to classify cell phases or identify UFBs.
- Apply the workflow to other cell types or biological structures.
- Expand analysis to include more advanced statistical comparisons across conditions.