The future of molecular diagnostics: divining the pathway

abstract AI data

This decade, like few others in memory, has underscored the importance of molecular diagnostics both to individuals and to populations. The emergence of SARS-CoV-2 initiated a global effort to understand the virus and detect its presence in asymptomatic patients, before it could be transmitted to others. With lessons learned from the pandemic, healthcare is now increasingly focusing on prediction and screening. Early detection and diagnosis of disease is crucial for the success of treatment pathways, particularly where late diagnosis could result in a condition developing beyond the point of intervention.

The recent case of young sisters Nala and Teddi Shaw highlights the importance of disease screening. Both sisters carry a mutation that causes metachromic leukodystrophy (MLD), a degenerative hereditary disease that results in nerve dysfunction from the accumulation of cerebroside sulfate in the myelin. The condition was not diagnosed until Nala, the elder sister, began to show symptoms at age 3. It is a tragedy that at the point that MLD symptoms present, the disease has already progressed too far to treat. Nala’s quality of life has already been affected irreparably and she was not expected to survive into adulthood. For Teddi, however, hope has come in the form of a revolutionary gene therapy, Libmeldy, developed by Orchard Therapeutics, UK. This is the most expensive treatment provided by the NHS in its history, but weighed against the cost of years of palliative care and of course the incalculable outcome for young Teddi, it is certainly worth pursuing.

Diagnostics in the lab

Omics: recent advances in molecular diagnostics

I will begin with a brief overview of ‘omics’, the science and technologies that underpin many of the advances in diagnostics over the last decade. ‘Omics’ are the study of molecular biological systems, whether genetic patterns (genomics), cellular protein expression (transcriptomics), the pathways of protein action (proteomics) or the chemical fingerprints of cellular processes (metabolomics). These fields in combination are called ‘multiomics’, which can provide a complete picture of the molecular markers of a disease.

Genomics

Genomics is the study of the entire genetic makeup of an organism, and involves analysing and interpreting the structure, function and organisation of genes within a genome. Genomics gives insights into how genes function individually and collectively, how they interact with the environment and how variations within the genome contribute to phenotypic differences. It is the most mature of the ‘omics’ and as a result has had the greatest development.

gene sequencing

The field of genomics has expanded rapidly in the last two decades thanks to the advent of next-generation DNA sequencing (NGS) technologies. These technologies automate the sequencing and analysis of DNA, providing high throughput data acquisition. Illumina is the market leader in NGS, whose technology is based on sequencing-by-synthesis of randomly digested short fragments in nanostructured flow cells. Their latest sequencer, the NovaSeq X, can run 128 genomes per day, at an approximate cost of $200 per genome, after capital expenditures. Illumina devices excel at producing high quality short-read data and has greater capacity for long-reads currently in development. Other molecular diagnostics technologies, such as the Sequel II from PacBio or the MinION and PromethION from Oxford Nanopore Technologies, specialise in much longer read-lengths, but with necessarily slower data acquisition times.

Transcriptomics

Transcriptomics is the study of the total set of expressed RNA in a cell, or population of cells, at a specific moment in time. Analysing a transcriptome provides a comprehensive understanding of which genes are active and producing RNA, as well as the quantity and variation of RNA molecules. This information helps in deciphering the underlying molecular mechanisms involved in various biological processes, such as development, disease progression and response to external stimuli. The mechanisms for the expression of DNA information into proteins are complex and highly regulated by processes not obvious from genomic information alone.

Transcriptomics is applied to the diagnosis and profiling of cancer through the identification of gene expression patterns associated with different types of cancer. It can also be applied to infectious diseases by analysing pathogenic transcriptomes or the host immune response and neurological disorders, by looking for expressed biomarkers in the brain or peripheral tissue.

The technology already developed for DNA sequencing is also capable of sequencing RNA with modifications to the biochemistry, called RNA-seq. The major suppliers of NGS devices and services also provide kits for RNA-seq.

Proteomics

Proteomics is the study of the entire set of proteins produced by a cell or tissue at a single point in time. It involves the identification, characterisation and quantification of proteins, as well as the study of their functions and interactions within biological systems. The approach can also identify post-translational modifications that effect protein function.

Proteomics uses molecular diagnostic techniques common across biotechnology and life science laboratories, rather than relying on a handful of devices as is the case for genomics and transcriptomics. Mass spectrometry (MS) is the principal technique to identify and characterise proteins. MS is not inherently quantitative, but labelling methods have been developed over the past decade to provide accurate quantification of protein expression. Quantitative proteomics is of particular importance when tracking disease progression and response to treatment over time. There has been a recent push towards protein nanopore sequencing, analogous to DNA nanopore sequencing. However, polypeptides are at least an order of magnitude more complex than DNA and so, present unresolved problems to sequencing. 

Metabolomics

Metabolomics is the study of metabolites, small molecules within cells, tissues or biological fluids that are the products of cellular processes. It is the youngest of the ‘omics’ but shows the most promise for molecular diagnostics and personalised medicine in clinical practice. An understanding of the complete set of metabolites within a biological sample provides insights into the biochemical pathways that underlie cellular function. Metabolite levels and interactions change in response to genetic makeup, environmental influences or drug treatments and so, can be excellent indicators of the presence of disease or the progression of treatment.

Metabolome biomarker discovery utilises common chemical analysis techniques, such as LC-MS and NMR, to characterise small molecule metabolites. Equipment of this kind requires large capital expenditure to purchase and significant expertise to operate, so metabolomic discovery is offered as services by research institutes or businesses, such as Biogenity and Metabolon.

The translation of metabolomics into clinical practice has been slow due to the differences in case-control studies used for discovery, versus the sample-control set testing used for diagnosis. Problems persist in the standardisation of methodologies between group studies and analytical approaches to gathered data, making accurate comparisons between different data sets difficult. This is especially problematic for personalised testing, which rely on robust data sets for comparison. Thus far, these concerns have prevented regulatory approval, with no FDA approved metabolomic tests currently available. This is, however, expected to change as the field matures and rigorous standards are adopted.

The future transition from omics research into clinical diagnostics will focus on the identification of biomarkers that indicate the presence or progress of a disease. It is not economically viable to undertake a full suite of omics assessments for each patient, in most instances, so specific tests must be developed for specific diseases. Having considered the advances in omics over the last decade, the following will consider some of the future trends in molecular diagnostics technology.

Liquid biopsy

Liquid biopsy is proclaimed to be the next big step in diagnostics. It functions by looking for biomarkers in bodily fluids, typically blood, saliva or urine. Biomarkers may be extracellular vesicles, circulating tumour cells, cell-free DNA or aberrant metabolites. While potentially applicable to a wide range of diseases, the vast majority of liquid biopsy development to date has focused on the detection of cancer. There is obvious benefit to analysing patient samples taken by minimally invasive means, instead of the surgical procedures required for traditional biopsies. The approach can not only be used as a tool for diagnosis, but also to monitor tumour burden and response to treatment.

The opportunity space is quickly being filled with ambitious start-ups taking multi-disciplinary approaches to the detection of ultra-low concentration biomarkers from liquid biopsy. For example, Cambridge-based Mursla is developing a platform for the detection and characterisation of extracellular vesicles from blood serum. Israeli start-up Nucleix is creating diagnostic tests for circulating tumour DNA from bladder and lung cancer. Meanwhile, Angle, UK, has built an FDA-cleared device for the capture and harvest of circulating tumour cells.

The push to realise the benefits from liquid biopsy has been huge, both from industry and academia, but the technology is still in its infancy. Current technology is not sufficiently robust to provide a complete diagnosis. The prevalence of false negatives and false positives is too high to confidently provide patients and clinicians with a definite yes/no answer to the question of cancer. Thus, there are still questions about the efficacy of liquid biopsy as a diagnostic tool compared to mature technology.

Galleri® is a multi-cancer early detection test from Grail, a spin-out from Illumina. Galleri® is currently undergoing clinical trial in the UK and is operating in the US under a CLIA waiver. The test screens for over 50 types of cancer, by sequencing cell-free DNA from a blood draw and identifying methylation patterns that indicate the presence and location of cancer.

Other companies are developing similar technologies. Exact SciencesFreenome and Guardant Health have all developed multiomics-based liquid biopsy tools and have reached various stages of clinical trial, including tens of thousands of participants. Despite this, no technology has full regulatory approval. Liquid biopsies are only in the first few steps of their journey to widespread adoption, but their future looks very promising.

Predictive and personalised molecular diagnostics

Rare diseases affect less than 6% of the UK population and can have debilitating consequences for patients that suffer from them. 80% of rare diseases are genetic and 50% manifest in childhood, but diagnosis can be extremely challenging due to the low number of patients and the diversity of presented symptoms. There are now many contemporary technologies for the early diagnosis of disease. The ‘holy grail approach’ is to identify and pre-emptively mitigate or treat disease for each individual patient and to determine the best treatment pathway with which to do so. Multiomics is incredibly powerful in this space, especially for complex heterogeneous diseases, where significant differences between patients exist.

Whole genome sequencing (WGS) is quickly becoming the gold standard in predictive medicine, providing explicit information about the risk of genetic diseases. WGS can identify the genetic markers of diseases within patent genomes before symptoms present or before the effects of a disease have begun to take hold. Timing is often of high importance, but there is evidence that a whole genome sequence can improve the diagnostic rate of genetic diseases by nearly a third, regardless of when it is performed.

However, a genetic predisposition to a disease does not guarantee that the disease will develop, nor is WGS alone sufficient to assess the spread of potential risk factors. For true molecular diagnostics, the full utilisation of multiomics is required for the identification of biomarkers associated with genotypic and phenotypic risk factors across the whole biology of a disease. This is especially the case for non-genetic disorders, where genetic propensity may be a risk factor, but not the cause in and of itself.

Quick and inexpensive biomarker screens can identify at-risk patients from large population groups for more targeted investigations. Screening for genetic biomarkers is currently the closest to widespread adoption because of the maturity of next-generation sequencing. However, screens for biomarker proteins and metabolites for screening disease such as Alzheimer’s and various mitochondrial disorders are also on the horizon.

Point-of-care diagnostic testing

The creation of rapid antigen lateral flow tests to detect SARS CoV-19 is one of the greatest diagnostic success stories, establishing point-of-care (PoC) and at-home testing within the cultural zeitgeist. Moving diagnostics from centralised laboratory facilities to the bedside (or near-bedside) reduces both the burden on health institutions and the time to results, benefitting both the patient and clinician. PoC in vitro diagnostic (IVD) tests are already common for many infectious and cardiovascular diseases, as well as for general health monitoring.

diagnostics test

There is a conceptual challenge in unifying personalised medicine and mass market PoC diagnostic devices, however. The molecular diagnosis of complex heterogeneous diseases requires the detection of multiple biomarkers, many of which depend on the specific biology of the individual patient. The requirement, therefore, is for a device to detect and assess biomarkers from a wide range of potential targets in a single test. This is beyond the capacity of most currently available simple PoC IVD tests, though devices that can detect multiple antigens do exist, such as the HemBox from Hememics Biotechnologies. It is expected that multiplexed PoC devices will be increasingly prevalent in years to come.

The creation of point of care diagnostics devices, or device platforms, is the ultimate objective of many businesses operating in the clinical diagnostics space, but transposing a diagnostic assay from laboratory to commercial device is not trivial. Manufacturing challenges can deeply affect the accuracy and reliability of PoC devices. Choices of labels, antibodies, biochemical materials and structural materials must be optimised for both device performance and the manufacturing process. Increasing device complexity compounds these challenges, so the path to manufacturing scale-up must be considered early in the development cycle in order to successfully launch such a device.

Data analysis, AI and diagnostics

All the technologies discussed in this article generate vast quantities of data. A single NovaSeq X can generate 8 Tb of data per day. It is impractical and undesirable for highly skilled clinicians and scientists to spend their valuable time sifting through this data. The adoption of artificial intelligence (AI) and machine learning (ML) into clinical and laboratory settings will help to automatically process and analyse patient data, to identify cases where further intervention may be necessary and to pick out the indicators of disease from multimodal sources. Indeed, the current major bottleneck in the widespread adoption of personalised diagnostic approaches is the integration of data from multiple sources for composite biomarker identification, especially for diseases that are not the result of solely genetic drivers.

abstract AI

On very large data sets, such as the EU 1+ Million Genomes Initiative, or the upcoming NHS Our Future Health project, AI represents an incredibly powerful tool to identify health trends on a whole population scale that would be next to impossible for researchers to detect. This represents a step change in preventative care for health services, where disease risk can be across ages, ethnicities and socio-economic groups.

Many businesses market their integration of AI or ML into their analytical services, but challenges remain. AI algorithms need to be trained on high quality and consistent data sets to make accurate evaluations, but medical data is often fragmented and incompleteAlgorithms also risk developing biases if not trained on sufficiently representative data sets. AI access to data from large healthcare initiatives helps to alleviate some of these issues, but there are on-going questions about privacy, data security and access to data for private enterprise.

The challenge of data security is certainly a source of some tension. AI analysis is increasingly powerful with increased access to data, but free access to sensitive biological and health information represents a real risk to patients and populations. The current consensus in the healthcare community is that population scale data should be stored in federalised trusted research environments (TREs), repositories where the data itself can only be accessed physically at that location, and access provided on a per-project basis. This, of course, slows the development of AI models for diagnostics, but is perhaps the best compromise for the protection and benefit of patients.

What does the future hold for molecular diagnostics?

The future of molecular diagnostics is one in which the risk of disease can be quantified on a per-patient bases, with diagnosis occurring at the point of care or in the home, taking into account the specific circumstances and biology of that patient. It is a future where powerful technologies provide faster, more accurate and less invasive testing. It is a future where, hopefully, disparities between patients in cases like Nala and Teddi Shaw can be eliminated.



Looking for something specific?