Overview
Explaining how genetic variation leads to disease means separating the few variants that disrupt function from the many that do not. We apply Omnii-1 to this problem across the genetic architecture of human disease, from closing the molecular diagnosis gap in rare disease to nominating candidate variants behind Parkinson's disease GWAS signals.
A central problem in human genetics is explaining how genetic variation leads to disease. At its core, this means separating the few variants that disrupt biological function from the many that do not. This task is especially hard for variants in non-coding DNA, which makes up the large majority of the genome. For a coding variant, the genetic code tells us how the protein changes: a variant that introduces a premature stop codon, for example, can be recognized as damaging from sequence alone. Non-coding variants, by contrast, often act by altering when, where, or how strongly a gene is expressed, and the gene they affect can lie far away along the genome. This makes them harder to interpret both experimentally, since testing them requires knowing which gene they regulate and in which cell type, and computationally, since existing tools have limited reach into these regions.
Genetic disease spans a range of settings (Figure 1). In Mendelian disease, one or two rare variants explain a patient’s diagnosis, and the challenge is to pinpoint them. In complex disease, common variants each have small effects, and the challenge is to identify the causal variant among the many correlated candidates at each genome-wide association study (GWAS) signal. Rare variants in complex disease fall in between: each is too rare to test alone, so the challenge is to decide which ones to analyze together. These settings call for different study designs and statistical tools, but they share a bottleneck: predicting from sequence which variants disrupt function, how they do so, and how that disruption connects to disease.
Figure 1. The genetic architecture of human disease. Disease-associated variants by allele frequency and effect size. Most lie along the diagonal band: rare alleles of large effect cause Mendelian disease, low-frequency variants of intermediate effect contribute to complex disease, and common variants of small effect are found by genome-wide association studies (GWAS). Each setting has its own study design, data and methods, but all share the same bottleneck. Adapted from Figure 2 of Manolio et al.1
Existing variant effect predictors each address a part of this question, but few extend reliably into non-coding DNA. Omnii-1, a multi-modal biological foundation model, is built to answer it directly, learning the regulatory code of DNA from sequence. Through state-of-the-art variant effect prediction encompassing both sequence-level constraint and molecular readouts, Omnii-1 can simultaneously find variants that exert the strongest effects and predict their functional genomic consequences. Moreover, as demonstrated in our Omnii preview, these predictions are not a function of minor allele frequency or local linkage disequilibrium (LD) structure, providing signals independent of the factors that typically confound inference in human genetics studies.
Here, we showcase Omnii-1's applications across the full genetic architecture of human disease, from Mendelian disease to complex disease, where we use Parkinson's disease (PD) to explore both rare and common variation. We present three vignettes (Figure 2):
1. Mendelian disease: closing the molecular diagnosis gap. In collaboration with Boston Children's Hospital, we apply Omnii-1 to five probands with congenital myopathy diagnoses but no full molecular diagnosis. The gap can be a causal variant that is non-coding, a variant of uncertain significance (VUS), or a missing second allele in a patient suspected to be compound heterozygous for a recessive gene. We show that Omnii-1 can nominate candidate causal variants that standard diagnostic pipelines have missed.
2. Rare variants in complex disease: Parkinson's age of onset. Age of onset in PD varies widely between patients, but carriers of rare predicted loss-of-function (pLoF) variants in a set of 21 canonical PD genes develop the disease about seven years earlier on average. This gives us a phenotypic readout to test against: if Omnii-1 is identifying truly disruptive variants, carriers of its top-scoring variants in these genes should show a similar shift toward earlier onset. In a large whole-genome sequencing (WGS) cohort of PD cases from the Global Parkinson's Genetics Program (GP2), we test this using gene burden tests. pLoF variants, which are defined by their consequence on the protein rather than by any model's prediction, establish the benchmark effect. We then ask whether each predictor's own top-ranked non-pLoF variants, coding and non-coding, recover a comparable shift, comparing Omnii-1 with widely used tools including CADD2, REVEL3, Borzoi4, Evo25, and AlphaGenome Variant Impact (AVI)6. We show that Omnii-1 matches the best existing predictors and finds disruptive variants more efficiently, recovering a larger shift in age of onset from fewer variants.
3. Common variants in complex disease: from benchmarks to Parkinson's GWAS. We approach common variants in two steps. First, we measure how often Omnii-1 selects the correct causal variant within GWAS and QTL credible sets, across a large collection of curated datasets where that variant has been nominated by functional experiments or statistical fine-mapping. This gives a broad view of how Omnii-1 performs when ground truth is available. We then apply Omnii-1 in the situation most researchers actually face, where there is no ground truth: PD GWAS in European, African, and South Asian ancestry cohorts, spanning 167 loci. Here, Omnii-1 nominates candidate causal variants, and its track predictions propose molecular mechanisms that could be tested in follow-up experiments.
Figure 2. Applying Omnii-1 across the genetic architecture of disease.
In these vignettes, ranging from finding a single causal variant in a patient's genome to ranking variants across hundreds of loci, we explore Omnii-1's predictions in real practice, demonstrating significant advancements over current tools.
Closing the molecular diagnosis gap in rare disease with Omnii-1
In collaboration with Catherine Brownstein, Alan Beggs, Katherine Anderson, and Casie Genetti of the Manton Center for Orphan Disease Research at Boston Children’s Hospital.
In rare or Mendelian disease, the goal is to find the one or two high-impact variants that explain a patient's diagnosis, usually by sequencing the patient and their family. Finding the precise molecular cause matters for families in several ways. It ends what is often a long diagnostic search, establishes the pattern of inheritance and the risk of recurrence in future children, can inform prognosis and gene-specific clinical care, and determines eligibility for gene-targeted trials and therapies. Typically, clinical laboratories classify candidate variants using guidelines from the American College of Medical Genetics and Genomics (ACMG)7, which weigh evidence such as population frequency, inheritance, functional data, and computational variant effect predictions. This framework leaves two major gaps. First, many variants lack enough evidence to be called either pathogenic or benign and are reported as variants of uncertain significance (VUS). Second, these guidelines are applied mostly to coding variants, but the causal variant may instead lie in non-coding DNA.
To further understand Omnii-1's capabilities to address these gaps, we utilized data from five unsolved cases of congenital myopathy obtained in collaboration with the Boston Children's Hospital Manton Center for Orphan Disease Research. Congenital myopathies are inherited muscle disorders that typically cause low muscle tone and weakness from birth or early childhood. We started with one trio: a girl with congenital fiber type disproportion (CFTD) and her unaffected parents. CFTD is diagnosed by muscle biopsy, in which one type of muscle fiber (type 1) is consistently smaller than the other. It can be caused by variants in several different genes, and many patients remain without a genetic diagnosis8.
All three trio members had long-read (Oxford Nanopore), short-read genome and short-read exome sequencing (Figure 3a). Having both parents meant we could assign every variant in the child to her mother or her father, screen for de novo variants, and use the long reads to place nearby variants on the same or opposite copies of a gene. We began with the genes already known to cause CFTD (TPM3, RYR1, ACTA1, MYH7, SELENON, TPM2 and LMNA)8. Across the three available sequencing platforms, 34 rare variants were called in the proband in this gene set, among which 10 were defined as well supported by the sequencing data (none de novo). Among these 10 candidates, the most compelling evidence emerged in RYR1, which maintains an autosomal recessive pattern of inheritance with CFTD, in which damaging variants on both inherited copies of a gene constitute a genetic diagnosis (Figure 3b).
Figure 3. The first trio and its three RYR1 variants. (A) Unaffected parents and their daughter with congenital fiber type disproportion (arrow), each sequenced with long reads (ONT), short-read exome (WES) and short-read genome (WGS). (B) The proband's rare, well-supported variants across RYR1 and its flanks, as drawn in the Omnii-1 Rare Disease Workbench: maternal above the gene, paternal below. Thick blocks are coding exons. Named are the maternal missense variant and the two paternal intronic variants. The unlabeled paternal variant lies downstream of the gene.
Methods
Search space
The genes to search were fixed from the diagnosis alone, before any variant was examined. For congenital fiber type disproportion these were the 7 genes listed by Clarke et al. 20118: TPM3, RYR1, ACTA1, MYH7, SELENON, TPM2 and LMNA. For nemaline myopathy they were the 13 genes listed by Sewry et al. 20199. Each gene's region spanned all of its GENCODE v47 transcripts, extended by 20 kb upstream and 10 kb downstream relative to the gene's strand, so that promoters and nearby regulatory sequence were searched along with the gene body.
Variant screening across platforms
All calls in these regions, from every platform a person had, were merged into a single table per person. Small variants (single-nucleotide changes, and insertions and deletions under 50 bp) were matched across platforms on chromosome, position and alleles after each allele was reduced to its minimal representation. Structural variants and copy-number changes (50 bp or larger) were merged across callers into a single event at 50% reciprocal overlap; insertions were merged if they lay within 500 bp of each other and were of similar length. A platform was counted as having found a variant only if its call passed the caller's own filters. Otherwise the table recorded whether that platform saw the reference allele at depth, did not cover the site (for example, outside the exome capture), or had no data.
Defining rare variants
A variant was considered rare if its allele frequency was at most 1% in both gnomAD v410 and CoLoRSdb v1.2.0. For gnomAD, the frequency used was the highest across all genetic ancestry groups, in both the exome and genome call sets. A group counted only if it had at least 1,000 alleles at the site, so that a small group at a poorly covered site could not make a variant appear common from one or two alleles. Structural and copy-number changes were compared against gnomAD-SV v4 and gnomAD's exome copy-number calls. CoLoRSdb, a set of 1,427 long-read genomes, was used with the same 1,000-allele minimum. It captures variation that short reads miss, particularly insertions and repeat-length changes, which can appear rare in gnomAD only because gnomAD cannot detect them. A variant absent from a reference set, or one where no group reached 1,000 alleles, was retained as rare, and the reason was recorded.
Call support
Low-confidence calls were flagged rather than removed, so every rare variant remained available for review. A rare variant was considered well supported if two or more platforms found it, or if the single platform that found it reported a genotype quality of at least 20. Structural variants followed the same rule using the caller's quality score, or at least 5 supporting reads for pbsv.
Inheritance, de novo variants and phasing
For each of the proband's rare variants, each parent's calls were examined on every platform that parent had. A parent was classed as a carrier if any platform called the variant, or if at least 3 reads, making up at least 15% of the depth, carried it without a call. A parent was classed as reference if no platform carried the variant and at least one platform observed only the reference allele at sufficient depth: at least 10 long reads or 15 short reads, with no more than one read carrying the variant. Otherwise the parent was classed as uncertain. A variant was then labeled maternal or paternal if only that parent carried it, inherited from both parents if both did, de novo if both parents were reference, and uncertain otherwise. Whether two variants in the same gene lay on the same copy (cis) or opposite copies (trans) was read from the proband's own phased long-read calls: two variants in the same phase block lay in trans if they were on different haplotypes and in cis if they were on the same haplotype. Where the parental calls could not place a variant, its long-read haplotype assigned it to a parent. This was done only within a phase block where at least two parentally placed variants agreed on that parent and none disagreed.
The maternal variant is a missense change, p.Pro3208Leu, already listed in ClinVar, where five of six submissions classify it as pathogenic or likely pathogenic and one as a VUS. Omnii-1’s DNA likelihood places it among the lowest-scoring 1.3% of rare variants genome-wide. Relative to other RYR1 missense variants in ClinVar, it scores lower than every benign or likely benign variant and lower than most pathogenic or likely pathogenic ones (Figure 4a). Omnii-1's evidence is therefore consistent with a pathogenic label for this variant, which constitutes a strong candidate first RYR1 hit.
However, the two paternal variants both lack entries in ClinVar and lie deep in introns. Genome-wide, Omnii-1 (DNA) placed c.801-724C>A in the lowest 9.7% of rare variants and c.6275-651G>A in the lowest 15.7%. Within RYR1, c.801-724C>A scores lower than 98.0% of the 754 rare intronic variants ClinVar classifies as benign or likely benign, and c.6275-651G>A lower than 91.8% of them (Figure 4b). Although c.801-724C>A scores as the stronger candidate against RYR1's benign variants, its position deep in an intron limits what the DNA score alone can tell us, so we turned to Omnii-1's track predictions to ask what each variant is predicted to change.
Figure 4. The three RYR1 candidates set among RYR1 variants in ClinVar of the same consequence, scored by Omnii-1 DNA. Rows are ClinVar classifications: pathogenic or likely pathogenic (P/LP), conflicting, uncertain significance (VUS), and benign or likely benign (B/LB). Rare variants (gnomAD at most 1%, or absent) are above each line and common ones below, each half scaled to its own tallest bar. (A) Missense variants, with the maternal variant marked. (B) Intronic variants, with the two paternal variants marked. The tables give the share of each group scoring as low as or lower than the family's variant. Their first row is the variant's percentile against rare variants across the genome.
We scored both paternal candidate variants with Omnii-1's predicted molecular tracks in skeletal muscle, spanning chromatin accessibility, CTCF binding, transcription initiation, histone marks, RNA expression and splicing (Figure 5a). Compared with the reference sequence, c.6275-651G>A was predicted to slightly lower two histone marks and RNA expression. While c.801-724C>A barely moved those tracks, it carried the most striking signal of either variant, a predicted new splice site not present in the reference sequence. It is also the most actionable one, because a new splice site is a specific mechanism that can disable a copy of the gene and can be tested directly using the patient’s RNA. Further, the variant sits 724 bases from the nearest exon in intron 9 and has never been seen in gnomAD. Omnii-1 predicts that it lifts splice site usage at a neighboring base from 0.003 to 0.18, about half the strength of the real splice sites at either end of the intron (Figure 5b). The sequence itself supports this, since the change turns GAG|GTCAGG into GAG|GTAAGG, a close match to the signal cells use to mark the end of an exon. We therefore propose a strong candidate for the second RYR1 hit, on the paternal copy and in trans with the maternal missense variant.
Figure 5. Omnii-1 track predictions for the two paternal RYR1 variants in skeletal muscle. (A) Top: predicted change in signal around each variant, as a percentage of the prediction for the reference sequence. Bottom: the strongest new splice site each variant is predicted to create, on a 0 to 1 usage scale because the reference value is near zero. The dashed line marks the real splice sites of intron 9. (B) Predicted splice site usage across RYR1 intron 9 with the reference sequence and with c.801-724C>A. Gray boxes are Alu repeats.
We then applied the same approach to the remaining four families, each with its own mix of available data from family members and sequencing platforms (Figure 6). We present the results in the Omnii-1 Rare Disease Workbench, a flexible framework for bringing Omnii-1 scoring into rare disease molecular diagnosis. For every family, the workbench:
1. Screens every variant class the available data can show, from SNVs to structural variants and copy-number changes, across whichever sequencing technologies a family has.
2. Places each variant on the mother's or the father's copy, using the parents where they were sequenced and long-read phasing where they were not.
3. Reads each variant's Omnii-1 scores, the DNA likelihood and the predicted tissue tracks, two ways: against rare variants across the genome, and against the gene's own ClinVar variants.
4. Ends with a hypothesis that grades the evidence for each copy of the gene.
Figure 6. The current hypothesis for each of the five congenital myopathy families, as summarized in the Omnii-1 Rare Disease Workbench. Family 1 is the trio described above. Sequencing data are long reads from Oxford Nanopore (ONT) or PacBio HiFi, short-read exome (WES) and short-read genome (WGS); grey means not present. The evidence for each of the gene's two copies is graded by hand against the rules in the key. Flags mark open issues.
Across all five families, the workbench points to a candidate diagnosis in four: NEB in the three families with nemaline myopathy, and RYR1 in the trio described above. In the fifth, a second family with CFTD, no candidate has yet emerged.
Through these explorations and the CFTD/RYR1 trio particularly (“Family 1”), we demonstrate the compelling utility of Omnii-1 in rare disease molecular diagnosis. We began with a child whose disease had no known genetic cause. Omnii-1 helped assemble a complete recessive diagnosis: it added evidence for the maternal missense variant, and on the paternal copy it singled out a deep intronic variant that standard interpretation would likely pass over. It also predicted what that variant does, creating a new splice site in skeletal muscle. That turns a candidate variant into a testable hypothesis, one that can be checked directly in the patient’s RNA. Like the other candidates, it remains a hypothesis until confirmed experimentally.
Acknowledgements
We thank the patients and families from the Manton Center congenital myopathy cohort, whose contribution of genomic and clinical information made this research possible.
Data used in the preparation of this blog were obtained from the Global Parkinson’s Genetics Program (GP2). GP2 is funded by the Aligning Science Against Parkinson’s (ASAP) initiative and implemented by The Michael J. Fox Foundation for Parkinson’s Research (www.gp2.org). For a complete list of GP2 members, see www.gp2.org.
Citation
@misc{omnii_genetics_2026,
title = {Applying Omnii-1 across the genetic architecture of human disease},
year = {2026},
month = {Oct},
url = {https://www.radicalnumerics.ai/blog/omnii-genetics},
organization = {Radical Numerics}
}References
1.Manolio, T.A., Collins, F.S., Cox, N.J., et al. "Finding the missing heritability of complex diseases." Nature 461, 747–753 (2009).
2.Schubach, M., Maass, T., Nazaretyan, L., et al. "CADD v1.7: using protein language models, regulatory CNNs and other nucleotide-level scores to improve genome-wide variant predictions." Nucleic Acids Research 52(D1), D1143–D1154 (2024).
3.Ioannidis, N.M., Rothstein, J.H., Pejaver, V., et al. "REVEL: An Ensemble Method for Predicting the Pathogenicity of Rare Missense Variants." American Journal of Human Genetics 99(4), 877–885 (2016).
4.Linder, J., Srivastava, D., Yuan, H., et al. "Predicting RNA-seq coverage from DNA sequence as a unifying model of gene regulation." Nature Genetics 57, 949–961 (2025).
5.Brixi, G., Durrant, M.G., Ku, J., et al. "Genome modelling and design across all domains of life with Evo 2." Nature 652, 1349–1361 (2026).
6.Cheng, J., Taylor, K.R., Nicolaisen, L., et al. "AlphaGenome Atlas: in silico mutagenesis of the entire human genome improves prioritization and interpretation of non-coding variants." medRxiv (2026).
7.Richards, S., Aziz, N., Bale, S., et al. "Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology." Genetics in Medicine 17(5), 405–424 (2015).
8.Clarke, N.F. "Congenital fiber-type disproportion." Seminars in Pediatric Neurology 18(4), 264–271 (2011).
9.Sewry, C.A., Laitila, J.M., Wallgren-Pettersson, C. "Nemaline myopathies: a current view." Journal of Muscle Research and Cell Motility 40(2), 111–126 (2019).
10.Guez, J., Goodrich, J.K., Moldovan, M.A., et al. "Integrating 730,947 exome sequences with clinical literature improves gene discovery." medRxiv (2026).
