75. EVALUATING THE IMPACT OF REFERENCE GENOME CHOICE ON PROMOTER TANDEM REPEAT GENOTYPING USING OXFORD NANOPORE TECHNOLOGIES LONG-READ SEQUENCING

Background Tandem repeats (TRs) in gene promoter regions modulate transcriptional regulation and have been implicated in schizophrenia (SCZ), bipolar disorder (BIP), and autism spectrum disorder (ASD), among others. As the field transitions from GRCh38 to telomere-to-telomere (T2T) references and long-read technologies make personalised de novo assembly feasible, researchers must choose between population-consensus and sample-matched references. It is unclear how this choice alters the TR landscape at gene promoters. Differences between assemblies in repetitive regions could bias TR calling, affecting the identification of clinically relevant regulatory variation. We systematically assessed reference-dependent TR annotation across human gene promoters, with a focus on psychiatric risk genes. We compared the current standard (GRCh38), a structurally complete reference (T2T-CHM13v2.0), and a de novo ONT genome assembly. Methods We generated a de novo assembly of the GIAB HG002 sample from 48x ONT reads using hifiasm (QV=52.4, BUSCO=98.6%). Protein-coding gene annotations were transferred via Liftoff, and a three-way matched promoter catalogue was constructed by identifying orthologous promoters (transcriptional start site+2kb/-0.5kb) across autosomal chromosomes from all three assemblies using MANE transcripts, flank identity, coordinate cross-liftover, and chromosomal and strand concordance. TRF was run with uniform parameters on promoter sequences ±1.5kb flanks to avoid truncating boundary-spanning repeats. TRs overlapping the original promoter region were retained for analysis. TR profiles were compared per gene, classifying promoters as concordant or discordant based on presence/absence, period, canonical motif, repeat count, and tract length. Psychiatric risk genes were identified from the NHGRI-EBI GWAS Catalog across five disorders (SCZ, BIP, ASD, major depressive disorder, and attention deficit/hyperactivity disorder). Results We established a catalogue of 17,997 three-way matched gene promoters, with a further 176 assembly-specific promoters present in only two of three assemblies. Across the matched set, 52% (n=9415) of promoters contained TRs in at least one assembly. Of TR-containing promoters, 38% (3528) showed discordant annotations across assemblies: period (1,221), presence/absence (983), count (595), tract length (374), and motif (355). Of 2,009 psychiatric risk genes in the catalogue, 392 exhibited discordant promoter TR profiles, including DRD2, RBFOX1, GRM8, NPAS3, ATXN7. Among assembly-specific promoter annotations, GRCh38 contributed the largest number of clinically relevant loci, including repeat-expansion disease genes DMPK and ATXN3. Discussion Reference assembly choice introduces variation in promoter TR genotyping, affecting about 40% of TR-containing promoters genome-wide, and nearly 400 psychiatric risk gene promoters. This discordance spans repeat presence, motif identity, and copy number, highlighting a silent confounder in TR-based association studies and emphasizing the need for reference-aware analytical frameworks. Ongoing work will extend this to sample-level genotyping by aligning ONT reads to each reference, generating polished consensus sequences, and applying TR genotyping tools to assess how reference choice impacts individual variant calls. We will further investigate how TR architecture influences genotyping accuracy and alignment quality.