The scale of the challenge becomes clear when one considers what whole-genome sequencing actually involves. Reading a person’s genetic information means capturing roughly 40,000 genes together with the vast regulatory regions that control their function. Dr. Gyula Richárd Nagy, a clinical geneticist and associate professor in the Department of Obstetrics and Gynecology at Semmelweis University, offers a vivid illustration of the volume: if all of this genetic text were laid out side by side, it would fill approximately 4,000 books of 500 pages each. The international Human Genome Project, launched in 1990 to map this enormous text, produced its first draft in 2001, but it was not until 2022 that researchers finally assembled a substantially complete, gap-free human genome. Even now, every new genome sequenced in a clinic or research laboratory must be interpreted by comparing it against a reference, and that comparison is performed almost entirely by automated software processing massive quantities of data.

The reference sequence currently in use worldwide is known as GRCh38. It represents the standard coordinate system against which variations in any individual’s genetic makeup are assessed. The problem, as the Semmelweis researchers point out, is that GRCh38 is not an average human genome that applies equally to everyone. It is built on a single, linear reference coordinate system, and although it incorporates genetic data from multiple individuals, roughly seventy percent of its content derives from the genome of a single male donor from Buffalo in the United States. When the software encounters a healthy genetic variant in a patient that happens to be listed in the reference as an exceptionally rare minor allele, it may trigger a false alarm. In other words, if the map fails to capture the true breadth of human genetic diversity, healthy traits can easily be mistaken for defects, particularly when the interpretation is left solely to automated pipelines without expert oversight.

The Hungarian study demonstrated exactly how this vulnerability plays out in practice. The whole genomes of twenty healthy volunteers were aligned against GRCh38, and in all twenty cases the automated analysis identified the same severe genetic variant. Only after expert curation, which cross-checked the findings against additional representative population regions, was the variant shown to be erroneous. None of the twenty individuals carried a medically relevant variant at all. Dr. Balázs Győrffy, Head of the Department of Bioinformatics at Semmelweis University and co-lead author of the paper, describes the result as the demonstration of an important bioinformatics vulnerability, one that arises not from faulty sequencing technology but from the inadequacy of the reference against which the sequences are judged.

At present, this weakness does not translate into clinical harm, because the discrepancies flagged by software are routinely reviewed and analyzed manually by specialists who consult other population databases before any conclusion is drawn. Human expertise acts as a safety net, catching the false positives that the automated systems generate. However, that safety net has a cost. Dr. Nagy warns that if whole-genome screening were expanded to large populations, the volume of manual corrections required would create an unsustainable bottleneck. Each false alarm demands expert time, and multiplying those demands across millions of screened individuals could overwhelm the very workforce needed to deliver the promised benefits of preventive genomics. In this sense, the reference genome problem is not merely a technical curiosity but a potential obstacle to the future expansion of population-wide preventive screening.

The solution proposed by the researchers is a graph-based pan-genome, a fundamentally different way of representing human genetic variation. Instead of describing the DNA sequence of a single person as one linear string, a pan-genome graph depicts the common and divergent segments of many human genomes in a branching, network-like structure. Such a computational reference already exists, and it offers a much richer genetic mapping system that covers human diversity more comprehensively. Crucially, it would eliminate the class of errors observed in the Hungarian study by design, because a variant common in some populations would appear in the graph as a known branch rather than as an alarming rarity. Comparing every person’s genome to a diverse network of genomes, rather than to a single reference sequence, provides a far more accurate picture of what is genuinely rare and what is simply part of normal human variation.

Yet the existence of a pan-genome graph does not mean it is ready for routine clinical use. Dr. Győrffy identifies the key question for the future as the extent to which information technology will be able to keep pace with advances in medicine. Current everyday IT systems, he notes, simply cannot support the daily use of such massive databases. Graph genomes are computationally far more demanding than linear references, requiring more storage, more processing power and more sophisticated alignment algorithms. Bridging the gap between a working computational prototype and a system that hospitals and screening programs can rely on every day is a formidable engineering challenge, and one that the researchers argue must be tackled proactively rather than reactively.

Dr. Nagy emphasizes that actively exploring the limits of the technology is essential if this branch of preventive medicine is to provide the highest level of safety without compromises. Genetic testing is already embedded in modern medicine in more targeted forms. For certain cancers, tumor samples are analyzed to identify genetic abnormalities that indicate which targeted therapy is likely to be effective. Genetic testing also plays an important role in diagnosing and treating rare conditions, in detecting hereditary diseases, and in assessing risk among affected family members, enabling early detection and preventive interventions in fields such as cardiogenetics. These applications demonstrate the clinical value of reading genetic information, but they operate on a far smaller scale than the whole-population vision that whole-genome sequencing promises.

The fundamental approach of future preventive and predictive medicine goes beyond treating disease once it has developed. The researchers argue that its core premise is knowing the risk before symptoms appear, and knowing who needs timely attention and for what. Whole-genome sequencing could give every person a personalized guide to their disease risks and predispositions, opening the door to genuinely individualized prevention strategies. That future, however, depends on a reference map worthy of the name. Until a graph-based pan-genome can be deployed routinely, and until the IT infrastructure catches up with the medical ambition, every automated genome analysis will carry the risk of mistaking healthy human diversity for disease, a risk that only expert human judgment currently stands between. The study, published in GeroScience under DOI 10.1007/s11357-026-02436-z, is a timely reminder that the tools we use to read the book of life must be as sophisticated as the book itself.

Subject of Research: Limitations of the GRCh38 human reference genome in automated whole-genome sequencing and the case for a graph-based pan-genome

Article Title: We may need a new map to properly interpret our genome

Article References: We may need a new map to properly interpret our genome. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: whole-genome sequencing, reference genome, GRCh38, pan-genome, bioinformatics, preventive medicine, genetic variants, false positives, personalized medicine, GeroScience, Semmelweis University, genomic screening

The CADD score works by training a machine-learning model to distinguish variants that have survived natural selection from those that have been purged from the human population over evolutionary time. Variants that look like the ones evolution tends to eliminate receive high scores, expressed on a PHRED-like scale where a score of 20 means the variant is predicted to be more deleterious than the top one percent of all possible mutations. In practice, laboratories often use a fixed cutoff, frequently around 20 or higher, to filter the flood of variants that emerge from whole-exome or whole-genome sequencing of a patient. Anything below the threshold is deprioritized; anything above it is flagged for closer scrutiny by a diagnostic team. The elegance of this approach is its simplicity. The problem, as the new paper makes clear, is that the simplicity is deceptive.

The distribution of CADD scores is not uniform across the genome. Different genes and different gene sets carry different baseline distributions of scores, meaning that a single global cutoff can behave very differently depending on the context. A threshold that works well for one disease panel may be too permissive for another, quietly letting potentially pathogenic variants slip through, or too strict, burying diagnosticians under false positives. Compounding the issue, scores can shift between genome builds, the two most common being GRCh37 and GRCh38, and between CADD releases, such as versions 1.6 and 1.7. A variant scored under one combination of build and release may land on a slightly different point of the scale when re-scored under another. For a field where a single misclassified variant can change a patient’s diagnosis, these are not trivial discrepancies.

CADD-ThresholdApp tackles the calibration problem by grounding threshold selection in real clinical evidence. The application draws on ClinVar, the public archive that collects and classifies variants according to their reported clinical significance, ranging from benign to pathogenic. By visualizing the CADD-score distributions of clinically classified variants, the app lets users see, for any gene or set of genes, how benign and pathogenic variants actually separate along the score axis. Rather than guessing where to place the cutoff, users can watch how the overlap between the two distributions changes as the threshold moves, and choose a boundary that reflects the genuine diagnostic trade-off between sensitivity and specificity in their specific context.

Beyond the visualizations, the app reports threshold-dependent performance metrics, quantifying how well a chosen cutoff performs at discriminating pathogenic from benign variants within the selected gene set. This turns what has often been an informal, intuition-driven decision into a measurable one. Users can define their own genes or gene lists, which is essential for laboratories working on rare diseases where only a handful of candidate genes matter. The app also integrates curated gene panels from PanelApp, a widely used resource of expert-reviewed gene panels for clinical genomics, allowing diagnosticians to calibrate thresholds directly against the panels they actually deploy in practice.

One of the most technically significant features of the application is its support for comparisons across genome releases and CADD versions. Because GRCh37 and GRCh38 differ in their coordinate systems and in some sequence content, and because successive CADD releases incorporate updated annotations and retrained models, the same variant can carry different scores in different pipelines. The app allows users to examine how these differences affect threshold behavior, which is particularly valuable for laboratories transitioning between genome builds or updating their scoring pipelines while maintaining consistency in variant interpretation. Being able to compare v1.6 and v1.7 side by side within the same interface means that a lab can assess the practical impact of an upgrade before committing to it.

The developers also emphasized extensibility. CADD-ThresholdApp is implemented in a modular framework, meaning that it can be adapted to incorporate additional variant resources beyond ClinVar and additional deleteriousness scores beyond CADD. This design choice matters because the variant-interpretation landscape is crowded with competing scores, including SpliceAI, REVEL, AlphaMissense, and many others, and laboratories increasingly combine multiple metrics. A tool that can grow to accommodate new evidence sources is far more durable than one hard-wired to a single score. The modular architecture also invites community contributions, allowing other groups to extend the app to their own scoring systems without rebuilding it from scratch.

Accessibility was clearly a priority. The application is available as a hosted web service at cadd-threshold.kircherlab.bihealth.org, requiring nothing more than a browser, which lowers the barrier for smaller diagnostic laboratories and research groups without dedicated bioinformatics support. For institutions with stricter data policies or custom requirements, the app can be installed locally via container images, from the Python Package Index through PyPI, through the Bioconda package manager, or directly from source code. This range of deployment options reflects a pragmatic understanding of how computational tools actually get adopted in clinical settings, where reproducibility and local control are often non-negotiable requirements.

The significance of this work extends beyond a single tool. Variant interpretation sits at the heart of modern genomic medicine, and the choice of scoring thresholds directly influences which patients receive diagnoses and which variants are dismissed as benign. As whole-genome sequencing becomes routine in healthcare systems worldwide, the volume of variants requiring classification is growing faster than the expert capacity to review them, making automated prioritization scores like CADD indispensable. But an indispensable tool calibrated carelessly is a liability. By providing a transparent, evidence-based framework for context-specific threshold calibration, CADD-ThresholdApp addresses a genuine blind spot in the diagnostic pipeline, one that has rarely been discussed openly because the field has defaulted to convenient universal cutoffs.

The study, published as open-access software research with no competing interests declared, represents a collaboration rooted in the lab that originally developed the CADD score itself, with Martin Kircher among the co-authors and acknowledgments extending to the broader CADD team and supporting bioinformatics units at the Berlin Institute of Health at Charité and the University of Washington. For a discipline built on the promise of precision, there is something fitting about a tool whose entire purpose is precision in calibration, ensuring that the numbers geneticists rely on mean the same thing, whether they are examining a cardiac gene panel in Berlin or a rare-disease exome in Tokyo. As genomic medicine scales up, tools like this one may prove to be the quiet infrastructure that keeps the entire diagnostic enterprise honest.

Subject of Research: Calibration of CADD variant deleteriousness score thresholds using ClinVar for clinical variant interpretation

Article Title: CADD-ThresholdApp: an interactive web application for calibrating CADD scores using ClinVar

Article References: Leifheit, C., Kircher, M., & Schubach, M. (2026). CADD-ThresholdApp: an interactive web application for calibrating CADD scores using ClinVar. BMC Bioinformatics, 27(1), Article 224. https://doi.org/10.1186/s12859-026-06675-x

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06675-x

Keywords: CADD score, ClinVar, variant interpretation, genomics, bioinformatics, gene panels, threshold calibration, rare disease, web application, GRCh38, variant prioritization, machine learning