Minimalist ensemble algorithms for genome-wide protein localization prediction

Abstract

BackgroundComputational prediction of protein subcellular localization can greatly help to elucidate its functions. Despite the existence of dozens of protein localization prediction algorithms, the prediction accuracy and coverage are still low. Several ensemble algorithms have been proposed to improve the prediction performance, which usually include as many as 10 or more individual localization algorithms. However, their performance is still limited by the running complexity and redundancy among individual prediction algorithms.ResultsThis paper proposed a novel method for rational design of minimalist ensemble algorithms for practical genome-wide protein subcellular localization prediction. The algorithm is based on combining a feature selection based filter and a logistic regression classifier. Using a novel concept of contribution scores, we analyzed issues of algorithm redundancy, consensus mistakes, and algorithm complementarity in designing ensemble algorithms. We applied the proposed minimalist logistic regression (LR) ensemble algorithm to two genome-wide datasets of Yeast and Human and compared its performance with current ensemble algorithms. Experimental results showed that the minimalist ensemble algorithm can achieve high prediction accuracy with only 1/3 to 1/2 of individual predictors of current ensemble algorithms, which greatly reduces computational complexity and running time. It was found that the high performance ensemble algorithms are usually composed of the predictors that together cover most of available features. Compared to the best individual predictor, our ensemble algorithm improved the prediction accuracy from AUC score of 0.558 to 0.707 for the Yeast dataset and from 0.628 to 0.646 for the Human dataset. Compared with popular weighted voting based ensemble algorithms, our classifier-based ensemble algorithms achieved much better performance without suffering from inclusion of too many individual predictors.ConclusionsWe proposed a method for rational design of minimalist ensemble algorithms using feature selection and classifiers. The proposed minimalist ensemble algorithm based on logistic regression can achieve equal or better prediction performance while using only half or one-third of individual predictors compared to other ensemble algorithms. The results also suggested that meta-predictors that take advantage of a variety of features by combining individual predictors tend to achieve the best performance. The LR ensemble server and related benchmark datasets are available at http://mleg.cse.sc.edu/LRensemble/cgi-bin/predict.cgi.

Predicting protein function and other biomedical characteristics with heterogeneous ensembles

“…This insight opens avenues for new algorithms for heterogeneous ensemble learning that can balance diversity and performance in a structured and mathematically rigorous way. Lin et al (2012) made a commendable beginning in this direction by proposing such an algorithm for enhancing the prediction of protein subcellular localization. However they applied these algorithms to a set of only about ten base predictors for this problem, and the true potential of these algorithms will be realized when they can be applied to the scale (number and type) of base predictors tested in our study.…”

“…Diversity measures, such as the Q statistic discussed earlier, might play an important role here. Approaches such as those proposed by Lin et al (2012) can be an important step forward in this direction. On the implementation front, although our DataSink framework exploits several parallelization opportunities to make the heterogeneous ensemble learning process more efficient, there are potentially many more avenues to improve this efficiency further.…”

The Meloidogyne graminicola effector Mg16820 is secreted in the apoplast and cytoplasm to suppress plant host defense responses

“…Prediction of cellular localization was performed using Wolf PSort (https://wolfpsort.hgc.jp) and LOCALIZER (https://localizer.csiro.au; Sperschneider et al, ). Prediction of the presence and localization of the NLS was performed with SeqNLS (Lin et al, ). The coding region of Mg16820 was blasted using default parameters with a cut‐off value of 50 (bitscore) against the M. graminicola genomics resource database (MGRAMBASE; https://insilico.iari.res.in/mgram/), the non‐redundant protein and nucleotide databases for all organisms of the National Center for Biotechnology Information (NCBI), fungal and oomycete genomes (FungiDB; https://fungidb.org/fungidb/), bacterial genomes (Microbial nucleotide blast on NCBI), and the four genomic datasets of the plant‐parasitic nematodes Bursaphelenchus xylophilus , Globodera pallida , M. incognita and M. hapla.…”

BIG3 Inhibits the Estrogen-Dependent Nuclear Translocation of PHB2 via Multiple Karyopherin-Alpha Proteins in Breast Cancer Cells

“…The binding capability of PHB2 with each KPNA was comparable with that of the Ala-replaced NLS mutant PHB2 (R86A, R88A and K89A) ( Fig 1C ). Notably, the scores of NLS sequences within PHB2 predicted using the SeqNLS algorithm [ 25 ] and cNLS Mapper [ 26 ] were substantially less than the cut-off value (data not shown). In recent years, however, it has become increasingly apparent that a number of proteins do not follow these canonical pathways and instead utilize non-conventional mechanisms [ 27 ].…”