The accurate prediction of star alleles of the CYP2D6 gene is crucial for pharmacogenomic precision, enabling tailored drug therapy based on individual genetic profiles. This study evaluates the efficacy of various bioinformatics algorithms for calling diplotypes, structural variants (SVs), and copy number variations (CNVs) of the CYP2D6 gene using whole genome sequencing (WGS) data from a cohort of 942 Thai individuals and validating our TPGxD-1 allele calls for CYP2D6. WGS data were processed using Illumina and HiSeq X technologies, with alignment to the GRCh38 reference genome. The Burrows-Wheeler Aligner and GATK tools were employed for read alignment, duplicate marking, and base quality score recalibration. Four algorithms—SGv2.0.2, sPGxv1.2.7, Aldyv4.0, and Cyrius v1.1.1—were used for star allele analysis. The results indicated that Aldyv4.0 had the highest calling rate for diplotypes/SVs/CNVs at 99.3%, followed by SGv2.0.2, while other algorithms showed varying efficiencies. Concordance analysis revealed high similarity in diplotype/SV/CNV frequencies between SGv2.0.2 and other tools, with Cyrius v1.1.1 showing the highest concordance in diplotype frequencies (r = 0.99) and sPGxv1.2.7 in SV/CNV frequencies (r = 0.99). Phenotype prediction was conducted using SGv2.0.2 and sPGxv1.2.7, showing high concordance in predicted metabolizer statuses. However, discrepancies were observed, primarily due to differing algorithmic capabilities in handling complex genetic variations. Unique and discordant diplotypes/SVs/CNVs were identified, highlighting areas for further tool optimization. Overall, this research contributes to the refinement of CYP2D6 allele calls, providing insights into the performance of contemporary bioinformatics tools and their implications for pharmacogenomic testing.