Dataset Information

Efficient use of unlabeled data for protein sequence classification: a comparative study.

ABSTRACT: BACKGROUND: Recent studies in computational primary protein sequence analysis have leveraged the power of unlabeled data. For example, predictive models based on string kernels trained on sequences known to belong to particular folds or superfamilies, the so-called labeled data set, can attain significantly improved accuracy if this data is supplemented with protein sequences that lack any class tags-the unlabeled data. In this study, we present a principled and biologically motivated computational framework that more effectively exploits the unlabeled data by only using the sequence regions that are more likely to be biologically relevant for better prediction accuracy. As overly-represented sequences in large uncurated databases may bias the estimation of computational models that rely on unlabeled data, we also propose a method to remove this bias and improve performance of the resulting classifiers. RESULTS: Combined with state-of-the-art string kernels, our proposed computational framework achieves very accurate semi-supervised protein remote fold and homology detection on three large unlabeled databases. It outperforms current state-of-the-art methods and exhibits significant reduction in running time. CONCLUSION: The unlabeled sequences used under the semi-supervised setting resemble the unpolished gemstones; when used as-is, they may carry unnecessary features and hence compromise the classification accuracy but once cut and polished, they improve the accuracy of the classifiers considerably.

SUBMITTER: Kuksa P

PROVIDER: S-EPMC2681072 | biostudies-literature | 2009

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Efficient use of unlabeled data for protein sequence classification: a comparative study.

Kuksa Pavel P Huang Pai-Hsi PH Pavlovic Vladimir V

BMC bioinformatics 20090429

<h4>Background</h4>Recent studies in computational primary protein sequence analysis have leveraged the power of unlabeled data. For example, predictive models based on string kernels trained on sequences known to belong to particular folds or superfamilies, the so-called labeled data set, can attain significantly improved accuracy if this data is supplemented with protein sequences that lack any class tags-the unlabeled data. In this study, we present a principled and biologically motivated com ...[more]

PMID: 19426450

Similar Datasets

Project description:Protein-protein interactions play a key role in many biological systems. High-throughput methods can directly detect the set of interacting proteins in yeast, but the results are often incomplete and exhibit high false-positive and false-negative rates. Recently, many different research groups independently suggested using supervised learning methods to integrate direct and indirect biological data sources for the protein interaction prediction task. However, the data sources, approaches, and implementations varied. Furthermore, the protein interaction prediction task itself can be subdivided into prediction of (1) physical interaction, (2) co-complex relationship, and (3) pathway co-membership. To investigate systematically the utility of different data sources and the way the data is encoded as features for predicting each of these types of protein interactions, we assembled a large set of biological features and varied their encoding for use in each of the three prediction tasks. Six different classifiers were used to assess the accuracy in predicting interactions, Random Forest (RF), RF similarity-based k-Nearest-Neighbor, Naïve Bayes, Decision Tree, Logistic Regression, and Support Vector Machine. For all classifiers, the three prediction tasks had different success rates, and co-complex prediction appears to be an easier task than the other two. Independently of prediction task, however, the RF classifier consistently ranked as one of the top two classifiers for all combinations of feature sets. Therefore, we used this classifier to study the importance of different biological datasets. First, we used the splitting function of the RF tree structure, the Gini index, to estimate feature importance. Second, we determined classification accuracy when only the top-ranking features were used as an input in the classifier. We find that the importance of different features depends on the specific prediction task and the way they are encoded. Strikingly, gene expression is consistently the most important feature for all three prediction tasks, while the protein interactions identified using the yeast-2-hybrid system were not among the top-ranking features under any condition.

Project description:Single-cell RNA sequencing (scRNA-seq) is an invaluable tool for profiling cells in complex tissues and dissecting activation states that lack well-defined surface protein expression. For immune cells, the transcriptomic profile captured by scRNA- seq cannot always identify cell states and subsets defined by conventional flow cytometry. Emerging technologies have enabled multimodal sequencing of single cells, such as paired sequencing of the transcriptome and surface proteome by CITE-seq, but integrating these high dimensional modalities for accurate cell type annotation remains a challenge in the field. Here, we describe a machine learning tool called MultiModal Classifier Hierarchy (MMoCHi) for the cell-type annotation of CITE-seq data. Our classifier involves several steps: 1) we use landmark registration to remove batch-related staining artifacts in CITE-Seq protein expression, 2) the user defines a hierarchy of classifications based on cell type similarity and ontology and provides markers (protein or gene expression) for the identification of ground truth populations within the dataset by threshold gating, 3) progressing through this user-defined hierarchy, we train a random forest classifier using all available modalities (surface proteome and transcriptome data), and 4) we use these forests to predict cell types across the entire dataset. Applying MMoCHi to CITE-seq data of immune cells isolated from eight distinct tissue sites of two human organ donors yields high-purity cell type annotations encompassing the broad array of immune cell states in the dataset. This includes T and B cell memory subsets, macrophages and monocytes, and natural killer cells, as well as rare populations of plasmacytoid dendritic cells, innate T cells, and innate lymphoid cell subsets. We validate the use of feature importances extracted from the classifier hierarchy to select robust genes for improved identification of T cell memory subsets by scRNA-seq. Together, MMoCHi provides a comprehensive system of tools for the batch-correction and cell- type annotation of CITE-seq data. Moreover, this tool provides flexibility in classification hierarchy design allowing for cell type annotations to reflect a researcher’s specific experimental design. This flexibility also renders MMoCHi readily extendable beyond immune cell annotation, and potentially adaptable to other sequencing modalities.

Dataset Information

Efficient use of unlabeled data for protein sequence classification: a comparative study.

Publications

Efficient use of unlabeled data for protein sequence classification: a comparative study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets