A comparison of categorisation algorithms for predicting the cellular localization sites of proteins

Abstract

A previous attempt to categorize yeast proteins based on certain attributes yielded only a 55% success rate of correct categorisation using a new type of decision procedure [4]. This paper considers using existing soft computing approaches to improve the categorisation. More specijically, learning algorithms based on neural networks, growing cell systems, a rule development algorithm and genetic algorithms are applied to the yeast data. All of the results are at least as good as the original datu showing that new problems do not necessarily require new algorithms. More interestingly, as a consequence of using diferent algorithms, a consistent failure to achieve high success rates actually indicates features of the data rather than the failings of one or other of the algorithms.

Analysing the localisation sites of proteins through neural networks ensembles

“…This behaviour provides evidence that a small variation in the value of the error goal might affect the classification success. This might provide explanation for the unsatisfactory FNNs performance reported in [10], where no details on the error goals used in the training phase were provided.…”

“…In order to explore the effect of the network error on the accuracy of the classifications, one hundred (100) independent trials were performed with two different termination criteria, which are based on the Mean Squared Error, E MSE <0.05 and E MSE <0.015. Following previous work in this area [6][7][8][9][10], we conducted experiments using all the data for testing, as well as data sets produced by fourfold cross-validation. Also, we did additional experiments by applying leave one out cross-validation to further investigate the performance of the methods.…”

“…The KNN achieved the best classification accuracy compared to these methods [9]: E. coli proteins were classified into eight classes with an average accuracy of 86%, while S. cerevisiae proteins were classified into ten classes with an average accuracy of 60% by applying cross-validation. More recently, attempts to improve the classification success have been made using back-propagation neural networks, genetic algorithms, growing cell structures and expanding range rules, but no significant improvements over the KNN algorithm were reported [10]. A data selection method for probabilistic neural networks was also applied to the E. coli dataset achieving better performance than the KNN (90% accuracy) [11].…”

Predicting Protein Localization Sites Using an Ensemble Self-Labeled Framework

“…In the past, there have been significant efforts for predicting the localization sites of proteins [18][19][20][21][22][23][24][25][26][27][28]. Anastasiadis and Magoulas [18] investigated the performance of K nearest neighbours, feedforward neural networks with and without cross-validation and ensemble-based techniques for the prediction of protein localization sites in E. coli and Yeast.…”

language impairment and colour categories

“…Moreover they are discrete in that they do not overlap. This could easily be changed by playing the game on standard categorisation tasks instead of colors (Cairns et al 2001). Human agents could even move the categories so that the system shared the vocabulary with the agents.…”