Background

Cancer is one of the leading causes of morbidity and mortality, with 10 million new cases being reported each year by the WHO. Despite the presence of several anti-cancer agents, cancer treatment is still not very effective due to the lack of tumor specific drug delivery systems. Most of the current anti-cancer drugs are unable to differentiate between cancerous and normal cells, leading to systemic toxicity, and adverse side effects. In order to address this problem, a considerable progress has been made over the years to identify peptides, which specifically bind to the tumor cells, and tumor vasculature which are collectively known as tumor homing peptides (THPs). With the advances in phage display technology, and libraries for screening peptides hundreds of THPs, have been discovered with the ability to specifically target tumors and deliver anti-cancer drugs to tumor sites. Currently, many tumor homing peptide-based therapies for cancer treatment and diagnosis are being tested in various phases of clinical trials.

THPep

The identification and subsequent development of THPs is time-consuming and expensive. Hence, it is extremely cost effective to develop a computational prediction method for the discrimination of THPs from non-THPs. In that regard, the proposed predictor, THPep, represents an effective and interpretable sequence based model using random forest method cooperated with amino acid composition, dipeptide composition and pseudo amino acid composition. In addition, 5-fold cross-validation and an independent testing sets are applied to the prediction results in order to obtain the most robust and accurate results.