Hydroponics research at Kelpinor

Generating robust evidence in biological research is not something that happens overnight. Good quality research requires patience with countless trials and errors, meticulous control over contamination to minimize errors, and painstaking data collection to ensure no relevant spatial or temporal points are missed(or, as we call it, embracing our inner nerds/control freaks). Thus began many days (and nights) of reading, planning, brainstorming (also known as arguing and nagging), coding, and not to mention fixing some (or many) frustrating leaks. With time and effort from the entire team, our plant research facility has made impressive progress. One of the early challenges was achieving accurate measurements with limited resources;

“Measuring a single plant won’t cut it. You need a lot of biological and technical replicates to create enough statistical power to support your hypotheses. But how are we supposed to measure the leaves of hundreds of plants every day?? Sacrifice our sleep? 3D print clones? Or just hire more interns (laughs nervously while crying)??”

The Solution

“HAIL MACHINE LEARNING!!!”

The pictures shows the workflow of an Machine learning model we developed. The analysis done by this model is integrated with an ImageJ macro to process image data collected daily. This automated setup measures leaf area across hundreds of plants, feeding the results into a Python-based statistical analysis pipeline for cleaning, exploratory data analysis, and hypothesis testing.

The results

The graphs in the video highlight findings from our preliminary experiments with mustard (Brassica juncea):

📈 Growth curves tracking total leaf area per day.

📊 A bar plot highlighting significant results.

As can be seen, certain concentrations of our product (ANS: Ascophyllum nodosum suspension) significantly improve leaf growth. While standardization remains a work in progress, we’ve made substantial strides. We’re now preparing for soil-based experiments with more plant species to further validate our findings and explore the full potential of our products.