If you want to be more precise, I would suggest to use INPARANOID to see their real ortholog sequence. INPARANOID is good in identifying real ortholog, inparalog etc.

Otherwise, you can choose the most significant one to annotate your gene. Somehow your Gene A and Gene B might be related to each other. And just my idea, you should see the alignment as well when you mention about the identity % as many most commonly used tools nowadays (like BLAST) use local aligment method whereby the identity % just reflect to the match of both sequences, not in overall of your gene.

thank u so much

We use CIP and CALP to calculate homologous and orthologus relationship between genes. (salse et al 2009) for more information see our paper in briefing of bininformatics title "Improved criteria and comparative genomics tool provide new insights into grass paleogenomics".

To better understand the evolutionary history of the sequences you are interested in, I would suggest you choose representative species from different taxa, search for orthologs or homologs with tools like INPARANOID, then check the MSA to see which regions are more conserved (which should be important for the protein's function, and whether there are sequences that have large insertions/deletions or might be false positives) and finally build a phylogenetic tree with PhyML, MrBayes or other available software, to see how the different sequences group and which are more closely related to each other.

OrthoMCL is a good choice, though usually more tricky to run than ProteinOrtho which is a nice option too.

You may like the recent PorthoDom.

To address your question, you can use OrthoMCL(http://sourceforge.net/projects/orthomcl/), which adoptes the Markov Clustering Algorithm (MCL) to perform a graph clustering of protein orthologs or close paralogous  from multiple eukaryotic genomes.

Enhua

The answer to your question is that they are both homologues. They might also both be orthologues of your gene of interest (i.e. diverged at a speciation event and subsequently duplicated), or neither might be an orthologue due to missing and/or lost genes. As @Aida suggests, the best way to ascertain this is to build a phylogenetic tree with these and additional homologues. If you need to do this at large scale, some of the automated predictors like INPARANOID are appropriate, though not as accurate as careful study.

If your example uses real numbers, I suspect that proteins A and B are both orthologues of X and paralogues of Y. i.e. There was an ancient duplication event (X&Y split), followed by a subsequent duplication in the X lineage (making A&B) after the two taxonomic groups in your studied diverged. (Y,(X,(A,B)));

thank you so much everyone for your valuable suggestion. took look further at the problem i read literature ant most Xs and Ys. it appears that X co ordinates in the physiologal functions performed by Y. after a definte phyllum in nimalia i see most  representative organisms from more recent phyllums have both X and Y. probably yes there was an ancient duplication event which gave rise to X and Y and later functional specialization occured for X and Y separately. thak you so much again.

however i was wondering if i could calculate the divergence time and evolutionary distances for my sequences. is there a way?

Phylogenetics again. I suggest finding yourself a good online tutorial using something reasonably accessible to the novice, like MEGA. To calculate a divergence (i.e. duplication) time you will need to have a fairly decent molecular clock in operation, which is rarely the case - especially in gene families. You also need to know the time of divergence of your taxa - the TimeTree website can help with this.