ParaDISM: New Tool Maps Tricky Short Reads to Lookalike Genes With Precision

Modern DNA sequencing machines generate billions of short reads, but a stubborn class of genes has long frustrated the software that places those reads on the reference genome. Genes that exist in multiple nearly identical copies — paralogs, tandem duplications and pseudogenes — are so similar in sequence that a 100- or 150-letter read may match several genomic locations equally well. When alignment software must choose among indistinguishable options, it can place reads on the wrong copy, and those misalignment artifacts ripple downstream through bioinformatic pipelines, inflating error rates in later steps such as single nucleotide variant calling. A team at the University of Malta has now built a pipeline designed specifically to confront this problem, and it is described in BMC Bioinformatics.

The tool, called ParaDISM — short for Paralog Disambiguating Mapper — refines standard alignments rather than replacing them. The core idea is to use a multiple sequence alignment of the reference sequences themselves to identify disambiguating positions: spots in the alignment where the highly homologous copies actually differ. A read or read pair is assigned to a particular gene copy only when the read carries sequence-specific evidence that unambiguously supports that copy. Reads that lack sufficient evidence for a gene of origin are reported as unassigned. This conservative strategy deliberately prioritizes specificity and precision — the positive predictive value of read placement — over sensitivity, accepting that some reads will go unplaced rather than risk placing them wrongly.

Why does this trade-off matter? In short-read high-throughput sequencing, the reads are so short that highly similar genomic copies create genuine ambiguity. Standard mappers such as Bowtie2, BWA-MEM2 and minimap2 use sophisticated scoring schemes to pick the best match, but when several locations tie, the choice is essentially arbitrary. Reads placed on the wrong paralog can carry variants from one gene copy into another, generating false positive variant calls. Those false positives can propagate through pipelines and, in clinical contexts, can lead researchers and clinicians to chase mutations that do not actually exist. ParaDISM attacks the problem at the alignment stage, before variant calling, by demanding stronger evidence before committing a read to a location.

The pipeline’s technical workflow is straightforward in concept. First, it takes the reference sequences for a region of interest — for example, a gene and its paralogs or pseudogene copies — and builds a multiple sequence alignment among them. From that alignment, it extracts the positions where the copies differ, the disambiguating positions. Then it examines each read: if the read spans or contains bases that match the sequence-specific pattern of one copy and contradict the patterns of the others, the read is confidently assigned to that copy. If the read’s evidence is insufficient — because it covers only regions where all copies are identical, for instance — it is left unassigned. The output is a refined BAM file, the standard Binary Alignment/Map format, so ParaDISM can slot into existing pipelines as a precision-focused add-on before the variant calling step.

An optional iterative refinement procedure addresses the sensitivity cost of the stringent placement criteria. In the first pass, ParaDISM calls variants from the confidently assigned reads. It then updates the reference sequences to incorporate sample-specific evidence — effectively personalizing the reference to the individual’s own variants — and re-processes the remaining unassigned reads. This gives the pipeline a second chance to place reads that were ambiguous against the generic reference but become unambiguous once the sample’s own sequence is known. The iterative loop mitigates the loss of sensitivity caused by the strict criteria for read placement while preserving the precision gains.

How well does it work? The team evaluated ParaDISM using extensive computational simulation experiments and the Genome in a Bottle HG002 benchmark, a widely used reference sample from the National Institute of Standards and Technology with a well-characterized genome. Performance was assessed both in terms of read alignment and the resulting short variant calls. Compared with Bowtie2, BWA-MEM2 and minimap2, ParaDISM reduced the number of misalignment artifacts, thereby increasing the specificity and precision of subsequent single nucleotide variant calls in highly homologous regions. The tool does not claim to outperform standard mappers everywhere; rather, it refines their output in the specific genomic contexts where homology makes their placements unreliable.

The researchers applied ParaDISM to two real-world case studies. The first involved five public tumour exomes at the GNAQ/GNAQP1 locus. GNAQ is a gene implicated in cancers including natural killer/T-cell lymphoma, and GNAQP1 is a pseudogene copy whose sequence similarity complicates variant interpretation. The second case study involved 18 short-read sequencing datasets from patients diagnosed with autosomal dominant polycystic kidney disease — 15 exomes and 3 targeted panel sequencing datasets. Autosomal dominant polycystic kidney disease is caused by variants in genes such as PKD1, which has a highly homologous pseudogene, making variant calling in this region notoriously difficult with short reads. ParaDISM was used to reanalyze these datasets, providing a stronger level of evidence for variants called in the homologous regions.

The clinical logic behind ParaDISM’s design is worth emphasizing. In applications where false positive variant calls have more severe consequences than false negative ones — such as identifying causal mutations for genetic disorders — a precision-focused tool is exactly what is needed. Missing a genuine variant is unfortunate, but reporting a variant that does not exist can trigger unnecessary follow-up, unnecessary wet-lab confirmation experiments, or even incorrect clinical conclusions. Where experimental confirmation of shortlisted variants is necessary, ParaDISM can be used to prioritize variants with strong support, reducing the cost and effort spent attempting to confirm false positive calls. The pipeline can also be run alongside the results of standard mappers to cross-check results and identify unreliable read placement and variants potentially caused by misalignment artifacts.

ParaDISM is open source, available under the MIT license at github.com/BioGeMT/ParaDISM, and its developers describe it as a refinement layer for BAM files rather than a replacement for standard mappers. The work emerged from a Maltese research collaboration spanning the Department of Applied Biomedical Science and the Centre for Molecular Medicine and BioBanking at the University of Malta, with funding from the Malta NGS project, the Genomics of Rare Diseases Project, and several European Union Horizon Europe programmes including Marie Skłodowska-Curie grants. The ADPKD study was approved by the University of Malta Research Ethics Committee, and written informed consent was obtained from all participants.

For the broader genomics community, ParaDISM represents a pragmatic response to a persistent technical challenge. As short-read sequencing remains the workhorse of clinical exome sequencing, panel sequencing and tumour genomics, genes with highly homologous copies will continue to generate ambiguous reads and unreliable variant calls. Tools that explicitly quantify and demand evidence for read placement — rather than silently accepting arbitrary choices among equally good matches — offer a path toward more trustworthy variant interpretation in these difficult regions. By refining alignments before variant calling and reporting unassigned reads transparently, ParaDISM gives analysts a clearer picture of where the evidence is strong and where it is not, and it invites pipelines to treat homologous regions with the caution they deserve.

Subject of Research: ParaDISM: precise mapping of short reads to genes with highly homologous regions

Article Title: ParaDISM: precise mapping of short reads to genes with highly homologous regions

Article References: Tzimotoudis, D., Farrugia, R., Zammit, J., Cini Masini, M., Balestrucci, A., Borg Carbott, F., Bezzina Wettinger, S., Alexiou, P., & Ciach, M. A. (2026). ParaDISM: precise mapping of short reads to genes with highly homologous regions. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06701-y

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06701-y

Keywords: ParaDISM, precise, mapping, short, reads, genes, highly, homologous, regions, scientific research