We introduce a new algorithm for the mass spectromet- ric identication of proteins. Experimental spectra obtained by tandem MS/MS are directly compared to theoretical spectra generated from pro- teins of evolutionarily closely related organisms. This work is motivated by the need of a method that allows the identication of proteins of unsequenced species against a database containing proteins of related organisms. The idea is that matching spectra of unknown peptides to very similar MS/MS spectra generated from this database of annotated proteins can lead to annotate unknown proteins. This process is similar to ortholog annotation in protein sequence databases. The difficulty with such an approach is that two similar peptides, even with just one mod- ication (i.e. insertion, deletion or substitution of one or several amino acid(s)) between them, usually generate very dissimilar spectra. In this paper, we present a new dynamic programming based algorithm: Packet- SpectralAlignment. Our algorithm is tolerant to modications and fully exploits two important properties that are usually not considered: the notion of inner symmetry, a relation linking pairs of spectrum peaks, and the notion of packet inside each spectrum to keep related peaks together. Our algorithm, PacketSpectralAlignment is then compared to SpectralAlignment [1] on a dataset of simulated spectra. Our tests show that PacketSpectralAlignment behaves better, in terms of results and execution time