MotifHider: A knowledge hiding approach to sequence masking

In a typical de novo motif discovery process, it is quite common that many of the motif candidates output from motif discovery programs are either already known motifs or motif-like decoy/repeat patterns. To prevent the false discovery and also to increase the chance of authentic novel motif discovery, some motif discovery programs employ a pre-processing stage in order to mask certain repeat positions in the input sequences. There are a few approaches to sequence masking aimed at avoiding the false discovery. This paper introduces a novel approach and a tool, called MotifHider, to sequence masking problem. MotifHider exploits sensitive knowledge hiding principles from database sharing. By hiding certain patterns, it provides successive motif discovery programs to avoid false discovery and rediscovery. At the same time, it avoids overly distortion of the input sequences so as to retain most of the authentic motifs.

I. Introduction

The problem of finding sequence motifs representing Transcription Factor Binding Sites (TFBSs) is an important challenge in bioinformatics, as these binding sites are the key to gene regulation. There are basically two approaches to binding site identification, experimental and computational. The experimental approach is precise but expensive and the computational approach is imprecise but inexpensive. Hence both approaches are viable and complementary. Significant portion of the current technological and scientific studies are aimed at to improve the drawbacks, i.e. towards the cheaper technologies for the former and the more accurate algorithms for the latter.