Structural annotation of genomes is one of major goals of genomics research. Most popular tools for structural annotation of genomes are determined by computational pipelines. It is well-known that these computational methods have a number of shortcomings including false identifications and incorrect identification of gene boundaries. Proteomic data can used to confirm the identification of genes identified by computational methods and correct mistakes. A Proteogenomic mapping method has been developed, which uses peptides identified from mass spectrometry for structural annotation of genomes. Spectra are matched against both a protein database and the genome database translated in all six reading frames. Those peptides that match the genome but not the protein database potentially represent novel protein coding genes, annotation errors. These short experimentally derived peptides are used to discover potential novel protein coding genes called expressed Protein Sequence Tags (ePSTs) by aligning the peptides to the genomic DNA and extending the translation in the 3' and 5' direction. In the paper, an enhanced pipeline, has been designed and developed for discovering and evaluating of potential novel protein coding genes: 1) a distance-based outlier detection method for validating peptides identified from MS/MS, 2) a proteogenomic mapping for discovery of potential novel protein coding genes, 3) collection of evidence from a number of sources and automatically evaluate potential novel protein coding genes by using machine learning techniques, such as Neural Network, Support Vector Machine, Naïve Bayes etc.