[Mississauga, Ontario] In recent years, the area of genetics has advanced significantly, particularly in the identification of previously unknown and rare disorders. However, the intricacy and quantity of genomic data have posed considerable problems. This blog post will explore the exciting developments in the automation of rare disease genetics data interpretation that leads to molecular diagnosis. We will draw insights from some key research highlighting the role of machine learning and advanced technologies in expediting the diagnosis of these debilitating conditions. Researchers [7] have indicated that automating the diagnosis of rare diseases through genome sequencing is a complex but promising endeavor that can significantly reduce the time and improve the accuracy of diagnosis.

Rare disease genetics and automation

According to the trusted Online Mendelian Inheritance of Man database (OMIM), out of the 7000 thousand rare diseases, there are over 6500 rare diseases that are genetic [8]. The number of these diseases is bound to increase as humans evolve and advances in medicine are made. Genetic diseases are adopted from either one of the parents or occur spontaneously later in life.

Due to the genetic variability of rare diseases, diagnosis is notoriously challenging, and exome and genome sequencing data interpretation can be time-consuming. Scientists [3] have even explored the challenges of diagnosing rare and undiagnosed diseases, emphasizing the need to explore alternative technologies when exome sequencing falls short. They suggest expanding the diagnostic toolkit to include whole-genome sequencing, long-read technology, pan-genome references, transcriptomics, metabolomics, proteomics, and methyl profiling. However, the main problem remains in the time and costs in manual interpretation of variants.

Multiple studies have shown enormous costs and lab power going to the process of manual interpretation of genetic variants. In one study published in an academic journal by the American college of medical genetics [14], the responses of 924 geneticists were collected and more than 92% were involved in clinical care. Half of these geneticists were operating with nearly full patient capacity. When questioned about their highest priority needs for patient management, multiple answers pointed towards technical assistance as the most significant factor needed. Another study [15] done in a UK national health laboratory sheds light on the comprehensive costs of sequencing and post-sequencing analysis, specifically in variant interpretation. From these different statistics, we can conclude that a lot of staff time and costs goes into the interpretation of genomes for molecular diagnosis. Not to mention that it usually takes almost three months to manually analyze the exome data in the clinic [9] and for the completion of this manual analysis, the time ranges from 5-10 years [13]. This increases diagnostic odyssey and patients can only receive symptom management during that time.

By automating the identification of potentially disease-causing genetic variants, clinicians can save valuable time and resources. It is especially crucial in the era of precision medicine, where timely diagnoses are essential for targeted treatments. 

The discovery of disease-causing genetic variations is being greatly sped up by cutting-edge sequencing technologies and artificial intelligence-driven algorithms [4]. These automated tools can quickly process enormous amounts of genomic data (for exome under five minutes in Horizon), assisting clinicians in fast determining the underlying cause of rare disorders. This discovery improves the probability of prompt interventions and better results while saving patients significant time [5].

One of these tools is Horizon, a clinical decision support tool designed to simplify and expedite genome interpretation. Horizon combines predictive methods based on patient symptoms, genome, and international guidelines (i.e. ACMG) enabling clinicians to nominate candidate diagnoses swiftly and accurately. With this cutting-edge technology, it will be easier to harness machine learning’s capacity to speed up the interpretation process, eventually assisting patients by cutting down on the time needed for diagnosis.

Integrating clinical data with genome

For a thorough diagnosis of a rare disease, the integration of clinical data and genetic information is essential. Automation is essential for effortlessly combining these two important diagnostic jigsaw pieces. Genomic data can be linked with patient medical histories, symptoms, pharmacogenetics databases and test findings by automated systems [3], giving clinicians a comprehensive understanding of the patient's condition. For example, Horizon platform integrates over thousand gene-to-drug information into their variant analysis pipeline. Because of this integration, diagnoses can be made more precisely and specifically, leading treatment choices based on genetic variables and clinical context [6].

Another crucial part is connecting the phenotype to these genetic findings. Clinical interpretation of genetic variants, particularly in the context of the patient's phenotype, can be a time-consuming and costly aspect of rare disease diagnosis. Artificial Intelligence can help with connecting phenotype to potential causative diseases and genes. Many platforms integrate human phenotype ontology (HPO) terms based on the phenotype. However, a further step should be taken in truly making sense of the data. One example is Horizon’s phenotype analysis module. Using natural language processing, Horizon interprets regular phenotype description from the user and provides the most probable diseases and genes causing this.

Through technology like this, automation becomes a standardized process that can enable clinicians and scientists to prevent, diagnose, and treat diseases.

Summing It Up

The automation of rare disease genetics interpretation is transforming the landscape of clinical genetics. Machine learning-based models and advanced technologies are expediting the diagnostic process and improving its accuracy. As highlighted above, integrating machine learning into clinical genetic laboratories holds tremendous promise for diagnosing rare diseases more swiftly and effectively. Researchers and clinicians are at the forefront of revolutionizing rare disease diagnosis by prioritizing candidate variants, nominating candidate diagnoses, and exploring advanced sequencing technologies beyond exome sequencing. These developments offer hope to millions worldwide who suffer from rare and undiagnosed diseases, promising quicker access to potentially life-changing treatments. The future of rare disease genetics interpretation is undoubtedly bright, driven by the power of automation and machine learning, propelling us toward a new era of precision medicine and improved patient outcomes.

References

1. https://doi.org/10.1186/s13073-02  1-00965-0
2. https://doi.org/10.1016/j.gim.2023.100830
3. https://doi.org/10.1186/s13073-022-01026-w
4. https://doi.org/10.1101/2022.12.07.22283238
5. https://ec.europa.eu/research-and-innovation/en/horizon-magazine/long-journey-rare-disease-diagnosis
6. https://doi.org/10.3389/fmolb.2023.1169109
7. https://doi.org/10.1007/s12687-020-00500-5
8. https://www.omim.org/statistics/geneMap
9. https://www.chop.edu/treatments/exome-sequencing#:~:text=Exome%20sequencing%20is%20a%20highly
10. https://doi.org/10.1016/j.semcancer.2021.06.009
11. https://doi.org/10.1038/s41551-022-00929-8
12. https://doi.org/10.1038/s41436-019-0566-2
13. https://doi.org/10.1186/s13023-020-01424-6
14. https://doi.org/10.1038/s41436-018-0417-6
15. https://doi.org/10.1038/s41436-019-0618-7