We developed an automated pipeline for the detection of single nucleotide polymorphisms (SNPs) in expressed sequence tag (EST) data sets, by combining three DNA sequence analysis programs: Phred, Phrap and PolyBayes. This application requires access to the individual electrophoregram traces. First, a reference set of 65 SNPs was obtained from the sequencing of 30 gametes in 13 maritime pine (Pinus pinaster Ait.) gene fragments (6671 bp), resulting in a frequency of 1 SNP every 102.6 bp. Second, parameters of the three programs were optimized in order to retrieve as many true SNPs, while keeping the rate of false positive as low as possible. Overall, the efficiency of detection of true SNPs was 83.1%. However, this rate varied largely as a function of the rare SNP allele frequency: down to 41% for rare SNP alleles (frequency < 10%), up to 98% for allele frequencies above 10%. Third, the detection method was applied to the 18498 assembled maritime pine (Pinus pinaster Ait.) ESTs, allowing to identify a total of 1400 candidate SNPs, in contigs containing between 4 and 20 sequence reads. These genetic resources, described for the first time in a forest tree species, were made available at http://www.pierroton.inra/genetics/Pinesnps. We also derived an analytical expression for the SNP detection probability as a function of the SNP allele frequency, the number of haploid genomes used to generate the EST sequence database, and the sample size of the contigs considered for SNP detection. The frequency of the SNP allele was shown to be the main factor influencing the probability of SNP detection.