An Efficient Mining Algorithm to Predict Transcription Factor DNA Binding Preferences

Transcription factor (TF) DNA binding preferences provide vast amounts of information about essential processes inside transcription regulatory mechanisms. Therefore, identifying DNA binding preferences of transcription factors is of prime importance. Several computational approaches have been proposed in previous works to develop a quick solution against the more expensive experiments. However, these computational approaches limit themselves to using existing biological information only while ignore the relationship between these properties. In this paper we take into account the weight and correlations of these biological properties and propose a novel computational approach, PWC (i.e., Predict TF DNA Binding preferences based on the Weight and Correlations of biological properties). By utilizing tf-idf method (term frequency inverse document frequency) to compute feature weight and GVSM (Generalized Vector Space Model) to compute feature correlations, PWC can provide a powerful approach to infer the TF DNA binding preferences. Our performance study on real data shows that PWC is better than other previous algorithms.