MEDEA clustering method is 'more effective'

Scientists have discovered a new process to improve the analysis of data compiled through Liquid Chromatography – Mass Spectrometry (LC-MS).

In a study published by BMC Bioinformatics, and undertaken by scientists from the Austrian Academy of Sciences, Harvard University and Harvard Medical School, a team sought to find a more effective and efficient clustering method.
Clustering is commonly used to recognise patterns in order to pinpoint groups of similar observations within a set of data.

The scientists used their new MEDEA (M-Estimator with DEterministic Annealing), an M-estimator based, new unsupervised algorithm that is designed to enforce position-specific constraints on variance during the clustering process on real-life LC-MS datasets.

As a result, they found that MEDEA not only out-performs the current state-of-the-art model-based clustering methods in identifying the common LC-MS peaks across multiple samples, it is also significantly more efficient, which means it can be used on larger datasets.

"MEDEA is an effective and efficient solution to the problem of peak matching in label-free LC-MS data," the study claimed.