Boosting multiclass learning with repeating codes and weak detectors for protein subcellular localization

Lin, Chung‐Chih; Tsai, Yuh-Show; Lin, Yutong; Chiu, Tai-Yu; Hsiung, Chia-Cheng; Lee, May-I.; Simpson, Jeremy C.; Hsu, Chun‐Nan

doi:10.1093/bioinformatics/btm497

Cited by 30 publications

(16 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Strong detectors are knowledge-driven features that are supposed to provide strong hints, while weak detectors are randomly extracted patterns to allow the learning algorithm to consider subtle characteristics of a class. This set of features has been used in (Lin et al , 2007) on recognizing fluorescent protein-tagged subcellular organelles in cell images, including mitochondria.…”

Section: Methodsmentioning

confidence: 99%

“…Recently, high-throughput image-based assay technologies, or high-content analysis, have become a useful tool for drug discovery (Jones et al , 2009; Lang et al , 2006), small molecule screen (Tanaka et al , 2005), subcellular localization (Huang and Murphy, 2004; Lin et al , 2007), etc. These technologies make it possible to visualize, trace and quantify cellular morphological changes in high resolution and play an increasingly crucial role to the understanding of biological processes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A spectral graph theoretic approach to quantification and calibration of collective morphological differences in cell images

Lin

Tsai

et al. 2010

Bioinformatics

Self Cite

View full text Add to dashboard Cite

Motivation: High-throughput image-based assay technologies can rapidly produce a large number of cell images for drug screening, but data analysis is still a major bottleneck that limits their utility. Quantifying a wide variety of morphological differences observed in cell images under different drug influences is still a challenging task because the result can be highly sensitive to sampling and noise.Results: We propose a graph-based approach to cell image analysis. We define graph transition energy to quantify morphological differences between image sets. A spectral graph theoretic regularization is applied to transform the feature space based on training examples of extremely different images to calibrate the quantification. Calibration is essential for a practical quantification method because we need to measure the confidence of the quantification. We applied our method to quantify the degree of partial fragmentation of mitochondria in collections of fluorescent cell images. We show that with transformation, the quantification can be more accurate and sensitive than that without transformation. We also show that our method outperforms competing methods, including neighbourhood component analysis and the multi-variate drug profiling method by Loo et al. We illustrate its utility with a study of Annonaceous acetogenins, a family of compounds with drug potential. Our result reveals that squamocin induces more fragmented mitochondria than muricin A.Availability: Mitochondrial cell images, their corresponding feature sets (SSLF and WSLF) and the source code of our proposed method are available at http://aiia.iis.sinica.edu.tw/.Contact: chunnan@iis.sinica.edu.twSupplementary information: Supplementary data are available at Bioinformatics online.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A spectral graph theoretic approach to quantification and calibration of collective morphological differences in cell images

Lin

Tsai

et al. 2010

Bioinformatics

Self Cite

View full text Add to dashboard Cite

show abstract

“…Since many other state-of-the-art approaches exalt the performance of a fusion of several descriptors, we expect to further improve our accuracy if other texture descriptors will be coupled with our idea. [22] 95.40 ----Hamilton et al [24] -98.20 93.20 --Lin et al [23] 93.60 ---- Table 12. Brief comparison between our results and other state of art descriptors and classifiers.…”

Section: Validation Of the Combination Of Global Feature Extraction Amentioning

confidence: 99%

“…But these results were obtained concatenating several descriptors. The best results obtained using a single descriptor were 90.8%, 95.7% and 91.8%;  in [22] a multi-resolution system obtains an accuracy of 95.4% in the HeLa dataset;  in [24] an ensemble of neural networks trained using the Haralick texture features and the Threshold Adjacency Statistics obtained the following accuracies in the END and TR datasets: 98.2 % and 93.2%;  in [32] the same descriptors were applied to HeLa, END and TR datasets and used to feed a random subspace of Levenberg-Marquardt neural networks with five hidden nodes, resulting in the following accuracies: 88.9%; 94.7% and 93.3%; in[23] where a novel AdaBoost method trained using global and local features obtains an accuracy of 93.6% in the HeLa dataset.…”

mentioning

confidence: 99%

Combination of projectors, standard texture descriptors and bag of features for classifying images

Nanni

Melucci

2016

Neurocomputing

View full text Add to dashboard Cite

“…Computational biology plays a vital role in the various areas of biology which involves pattern matching as a vital role in bioinformatics. Discrimination of cancer [2] evaluation from gene expression, protein sub-cellular localization from experimental data [3] by extracting features from raw images with possibility of applying genetic interactions to predict pathways have been evolved in this field.…”

Section: Introductionmentioning

confidence: 99%

Quantitative Based Quality Evaluation for Pattern Based Bioinformatics Tools

Manicassamy¹,

Dhavachelvan²

2009

IJCEE

View full text Add to dashboard Cite

-Today computational biology has merged with few other sciences for solving biological problems where there is a tremendous increment in innovations of new techniques and approaches that have been done day to day. In this field, area of research involves disease identification, sequence analysis, structural analysis, protein-protein interaction etc… in which pattern matching plays a major role. There are various tools that have been developed in pattern matching in the area of bioinformatics for various analysis processes. At present bioinformatics tools are increasingly important to allow real-time decision making for solving biological problems for which performance plays the key role for evaluating the developed tools. Performance of a tool can be evaluated by means both qualitative and quantitative which has been involved in this paper. Quantitative based evaluation represents the quality of the tools for which metric based evaluation like specificity, sensitivity, pattern match etc… are been carried out. Qualitative based evaluation represents features, functionality evaluation together with the performance quantitative assessment. The merit that we are discovering by evaluating the existing products, leads us in the development of new products by avoiding the pitfalls of the existing products. In this work, some of the pattern tools are been considered for evaluating the performance which involves standards evaluation procedure. The results observed during this evaluations made on the specified tools have made an indication for having a standard development cum evaluation of pattern based bioinformatics tools. These evaluations could provide a way for developing and maintaining better systems with enhanced qualities.

show abstract

Boosting multiclass learning with repeating codes and weak detectors for protein subcellular localization

Abstract: CHO and Vero cell images, their corresponding feature sets (SSLF and WSLF), our new learning algorithm, AdaBoost.ERC, and Supplementary Material are available at http://aiia.iis.sinica.edu.tw/

Cited by 30 publications

References 25 publications

A spectral graph theoretic approach to quantification and calibration of collective morphological differences in cell images

A spectral graph theoretic approach to quantification and calibration of collective morphological differences in cell images

Combination of projectors, standard texture descriptors and bag of features for classifying images

Quantitative Based Quality Evaluation for Pattern Based Bioinformatics Tools

Contact Info

Product

Resources

About