Investigating the pathogenicity of missense spliceogenic variants using whole-genome sequencing, RNA-sequencing, and proteomics data from population and clinical research cohorts

Human health relies on a complex instruction manual written in our DNA, a code that tells our cells how to build the proteins necessary for life. For decades, scientists have focused on how changes in this code alter the shape and function of the final protein, much like checking if a typo in a recipe ruins the taste of a cake. However, the instruction manual has a critical editing step before the final product is made. This process, known as splicing, cuts out unnecessary sections and stitches the remaining parts together to form a clean, usable message. If this editing step goes wrong, the cell may discard the message entirely or produce a broken version, leading to disease. While researchers have long known that errors at the very edges of these sections can cause such mistakes, they have largely assumed that changes in the middle of the sections are safe, provided the final protein still looks correct.

A new study challenges this assumption by revealing that many genetic changes previously thought to be harmless actually sabotage the editing process. By analyzing genetic data from nearly half a million people and measuring the levels of proteins in their blood, researchers discovered that a significant number of these "middle" changes disrupt the splicing machinery. These errors, which the team calls missense spliceogenic variants, often go unnoticed by current diagnostic tools because those tools only look at the final protein shape, not the editing process. The findings suggest that a substantial portion of genetic variants currently labeled as uncertain or benign are actually causing disease by preventing the cell from making the necessary proteins at all.

To understand the scale of this hidden problem, the research team turned to the UK Biobank, a massive database containing genetic information and health records from approximately 500,000 volunteers. They focused on a specific type of genetic change called a missense variant, which is a single letter change in the DNA code that usually results in a different building block being used in a protein. The team examined over eight million of these variants to see if any of them also disrupted the splicing process. They used a sophisticated computer model to predict how each variant would affect the editing of the genetic message. The results were striking: about 2.7 percent of these missense variants were predicted to interfere with splicing, a rate nearly identical to that of synonymous variants, which are changes that do not alter the protein at all. This suggests that the risk of disrupting the editing process is just as common in the middle of the code as it is at the edges.

The study then investigated whether these predicted errors actually had real-world consequences. The researchers compared the genetic data with blood samples from over 50,000 participants, looking at the levels of nearly 3,000 different proteins circulating in the blood. They found that individuals carrying these specific missense variants had significantly lower levels of the corresponding proteins. In many cases, the reduction in protein levels was as severe as that seen in people with known, devastating genetic errors that completely stop protein production. This confirmed that the predicted splicing errors were not just theoretical; they were actively preventing the body from making enough of the necessary proteins.

A critical part of the investigation involved testing the reliability of the tools currently used by doctors and geneticists to interpret these variants. Most standard software programs are designed to predict whether a genetic change will damage the protein's structure. The researchers found that for the variants they identified as splicing disruptors, these standard tools often gave a "safe" rating. In fact, more than half of the variants that were predicted to break the splicing process were classified as benign or neutral by the protein-focused tools. This creates a dangerous blind spot: a variant can be flagged as harmless because the protein looks fine, while in reality, the cell is failing to produce the protein in the first place. This discrepancy was even more pronounced in regions of proteins that are naturally flexible and lack a rigid structure, where standard tools are least likely to predict a problem.

To prove that their computer predictions matched biological reality, the team looked at actual genetic data from patients with rare diseases who had previously received no diagnosis. They identified 14 missense variants of uncertain significance from the NHS Genomic Medicine Service that their models predicted would disrupt splicing. When the researchers examined the RNA, the genetic message, from these patients, they confirmed that the splicing was indeed disrupted in these cases, providing strong evidence to classify the variants as pathogenic or likely pathogenic. In one instance, a variant in a gene called VRK1, which was previously thought to be harmless, was shown to cause the cell to skip an entire section of the message, leading to a severe neurological condition. In another case, a variant in the PRG4 gene, which causes joint and heart problems, was found to force the cell to use a wrong editing site, deleting a crucial piece of the message. These findings provided direct evidence that the computer models were correctly identifying disease-causing errors that had been missed by traditional methods.

The researchers also applied their new approach to a large group of unsolved cases from the 100,000 Genomes Project, a national initiative to find genetic causes for rare diseases. By screening for these specific types of splicing errors, they identified a strong diagnostic candidate in a patient who had been undiagnosed for years. The patient had a genetic change in a gene called DYSF that was previously considered uncertain. The new analysis revealed that this change destroyed the editing site, leading to a form of muscular dystrophy, and the findings strongly matched the patient's symptoms, warranting further clinical examination. They also found strong candidates for two other patients with similar unexplained conditions. These successes demonstrate that looking at the editing process, rather than just the final protein, can solve medical mysteries that have persisted for a long time.

The study concludes that the current way of interpreting genetic changes needs to evolve. For years, the focus has been on whether a change breaks the protein's shape, but this research shows that a change can be just as harmful if it stops the message from being edited correctly. By integrating tools that predict splicing errors into standard diagnostic workflows, doctors can uncover the true cause of disease in many more patients. This shift in perspective moves the field from a narrow focus on the final product to a broader understanding of the entire manufacturing process, ensuring that no genetic error is overlooked simply because it hides in plain sight.