Fig. 2.

Comparison of de novo sequencing tools [as well as a database search tool MS-GFDB (Kim et al., 2010) tweaked for de novo sequencing]. Per each spectrum, N top scoring reconstructions were generated by UniNovo, PepNovo+ (Frank, 2009; Frank and Pevzner, 2005), PEAKS (Ma et al., 2003), pNovo (Chi et al., 2010) and MS-GFDBScore. MS-GFDBScore provides UniNovo with MS-GFDB’s scoring function. The number of reported reconstructions per a spectrum (N) is set to 1, 5 and 20. A reconstruction is correct if all the fragmentation sites of the reconstruction are correct, and a spectrum is classified as correctly sequenced if at least one of the reconstructions generated from the spectrum is correct. Figures on the left side (a, c and e) show the number of correctly sequenced spectra in each dataset, and figures on the right side (b, d and f) show the average length of the correct reconstructions