Dataset Information

A regression-based K nearest neighbor algorithm for gene function prediction from heterogeneous data.

ABSTRACT:

Background

As a variety of functional genomic and proteomic techniques become available, there is an increasing need for functional analysis methodologies that integrate heterogeneous data sources.

Methods

In this paper, we address this issue by proposing a general framework for gene function prediction based on the k-nearest-neighbor (KNN) algorithm. The choice of KNN is motivated by its simplicity, flexibility to incorporate different data types and adaptability to irregular feature spaces. A weakness of traditional KNN methods, especially when handling heterogeneous data, is that performance is subject to the often ad hoc choice of similarity metric. To address this weakness, we apply regression methods to infer a similarity metric as a weighted combination of a set of base similarity measures, which helps to locate the neighbors that are most likely to be in the same class as the target gene. We also suggest a novel voting scheme to generate confidence scores that estimate the accuracy of predictions. The method gracefully extends to multi-way classification problems.

Results

We apply this technique to gene function prediction according to three well-known Escherichia coli classification schemes suggested by biologists, using information derived from microarray and genome sequencing data. We demonstrate that our algorithm dramatically outperforms the naive KNN methods and is competitive with support vector machine (SVM) algorithms for integrating heterogenous data. We also show that by combining different data sources, prediction accuracy can improve significantly

Conclusion

Our extension of KNN with automatic feature weighting, multi-class prediction, and probabilistic inference, enhance prediction accuracy significantly while remaining efficient, intuitive and flexible. This general framework can also be applied to similar classification problems involving heterogeneous datasets.

SUBMITTER: Yao Z

PROVIDER: S-EPMC1810312 | biostudies-literature | 2006 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A regression-based K nearest neighbor algorithm for gene function prediction from heterogeneous data.

Yao Zizhen Z Ruzzo Walter L WL

BMC bioinformatics 20060320

<h4>Background</h4>As a variety of functional genomic and proteomic techniques become available, there is an increasing need for functional analysis methodologies that integrate heterogeneous data sources.<h4>Methods</h4>In this paper, we address this issue by proposing a general framework for gene function prediction based on the k-nearest-neighbor (KNN) algorithm. The choice of KNN is motivated by its simplicity, flexibility to incorporate different data types and adaptability to irregular fea ...[more]

PMID: 16723004

Similar Datasets

Project description:BackgroundThe enumeration of tetrameric and other sequence motifs that are positively or negatively correlated with in vivo antisense DNA effects has been a useful addition to the arsenal of information needed to predict effective targets for antisense DNA control of gene expression. Such retrospective information derived from in vivo cellular experiments characterizes aspects of the sequence dependence of antisense inhibition that are not predicted by nearest-neighbor (NN) thermodynamic parameters derived from in vitro experiments. However, quantitation of the antisense contributions of motifs is problematic, since individual motifs are not isolated from the effects of neighboring nucleotides, and motifs may be overlapping. These problems are circumvented by a next-nearest-neighbor (NNN) analysis of antisense DNA effects in which the overlapping nature of nearest-neighbors is taken into account.ResultsNext-nearest-neighbor triplet combinations of nucleotides are the simplest that include overlapping sequence effects and therefore can encompass interactions beyond those of nearest neighbors. We used singular value decomposition (SVD) to fit experimental data from our laboratory in which phosphorothioate-modified antisense DNAs (S-DNAs) 20 nucleotides long were used to inhibit cellular protein expression in 112 experiments involving four gene targets and two cell lines. Data were fitted using a NNN model, neglecting end effects, to derive NNN inhibition parameters that could be combined to give parameters for a set of 49 sequences that represents the inhibitory effects of all possible overlapping triplet interactions in the cellular targets of these antisense S-DNAs. We also show that parameters to describe subsets of the data, such as the mRNAs being targeted and the cell lines used, can be included in such a derivation. While NNN triplet parameters provided an adequate model to fit our data, NN doublet parameters did not.ConclusionsThe methodology presented illustrates how NNN antisense inhibitory information can be derived from in vivo cellular experiments. Subsequent calculations of the antisense inhibitory parameters for any mRNA target sequence automatically take into account the effects of all possible overlapping combinations of nearest-neighbors in the sequence. This procedure is more robust than the tallying of tetrameric motifs that have positive or negative antisense effects. The specific parameters derived in this work are limited in their applicability by the relatively small database of experiments that was used in their derivation.

Dataset Information

A regression-based K nearest neighbor algorithm for gene function prediction from heterogeneous data.

Background

Methods

Results

Conclusion

Publications

A regression-based K nearest neighbor algorithm for gene function prediction from heterogeneous data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets