Today there continues to be no consensus on the classification of brain cells. Under the microscope there is a broad diversity of brain cell sizes and shapes. Beginning with the classical studies of Ramon y Cajal it’s been suggested that a relationship exists between brain cell shape and function.

Brain cell classification based on shape has been largely subjective. Recording electrical properties of identified cells has helped to quantify shape and function. Nevertheless, classification schemes vary between the “lumpers” (few different categories) and the “spliters” (many different categories) and everything in between.

A recent paper attempts to address brain cell classification based on shape by using cell data from the NeuroMorpho online data repository and clustering methods.

(The paper is titled “Investigating the Morphological Categories in the NeuroMorpho Database by Using Superparamagnetic Clustering” by Krissia Zawadzki and colleagues published in the March 15, 2010 issue of arXiv.)

NeuroMorpho is currently the largest online repository of the digital geometric representations of actual brain cells. This repository is an exciting start that could grow many orders of magnitude larger if every scientist is compelled to submit their morphology (cell shape) data at the same time they publish the results in a traditional journal.

The analysis reported in this paper explored brain cell shape measurements from NeuroMorpho using a classification procedure of statistical physics known as Superparamagnetic Clustering (SPC). SPC is an unsupervised method for clustering. They also used Principal Component Analysis (PCA) and Latent Dirichlet Allocation (LDA).

Their main objective was to compare the obtained clusters with the original classification in the repository and check the agreement with the original categories. My best assessment of their results leads me to conclude that the study was at best inconclusive.

In my view, studies like this are important for exploring what may or may not provide useful results when analyzing data from large repositories like NeuroMorpho. Publishing negative results is as important as publishing positive results. Let’s hope scientists will publish more of their negative results and do so without necessarily writing them up to sound like positive results.