Research Goal: Attain New Levels
of Science through Algorithms

The students in New Rochelle High School’s Science Research Program explore a dazzling range of topics under the mentorship of experts from some of the top institutions of learning and exploration. 



Student: Nikhil Mukrajjunior


Mentor: Anthony Nesturi, Formulation Development Specialist, Regeneron Pharmaceuticals



Nikhil Mukraj always has been fascinated by science and, as he says, how it “builds on itself.” He looks at science and math as a unique, intricate puzzle.



“Everything in science and math is based upon previous rules and then expanded into more and more complex concepts to explain all of the natural world,” he said. “If you understand the fundamental rules and logic of the game of science, then you are able to manipulate them into even more complex machines for your use, which then may in turn create more science to be discovered.”


This way of looking at science and math is what prompted Mukraj to get involved in the Science Research Program at New Rochelle High School.

Mukraj is working on a project in which he tested various algorithms to see how efficiently they can identify protein and silicon particles from data in images. The images are taken using an electron microscope. Protein particles can have harmful side effects in some drugs, so being able to quickly identify them could be greatly beneficial for the pharmaceutical industry.


Consistent with his interest in scientific rules and logic, Mukraj used machine-learning algorithms to find out which could identify proteins in images.


He discovered he could achieve the greatest accuracy with a convolutional neural network, a type of algorithm that takes an image, assigns importance to objects in, or aspects of, the image and differentiates them from each other. However, he also found he could get data faster with a support-vector machine, a type of machine-learning algorithm that learns by example to label certain types of objects. He believes the support-vector machine method could be used as a good baseline for future particle-classification algorithms.


Mukraj is still working on the project. His next step will be to redo the convolutional neural-network analysis with a novel technique that combines meta data and image data to derive a new way of evaluating results. 


Mukraj hopes to be able to continue his research after high school. For now, though, the Science Research Program continues to feed his interest in the field.


“Personally, the most rewarding aspect for me is researching in depth into one specific topic and actually being able to add research to that area,” he said.