The value of Omics biomarkers in clinical trials
Milan Geybels (Novo Nordisk), Carsten Henneges (Syneos Health), Andrejus Parfionovas (Bayer) and Julie Russell (Bayer)
What is Omics?
An ome is a totality. In molecular biology, the suffix -omics is used to represent comprehensive biological data such as proteomics (e.g., the expression of thousands of proteins in plasma), transcriptomics (e.g., the expression of tens of thousands of transcripts/genes in liver tissue), genomics (e.g., variation in your DNA). In addition, there is also the microbiome (i.e., all microbiota residing in your gut). As such, Omics datasets can be large. For example, a typical epigenomics dataset may include more than a million features per sample, which are called DNA methylation markers.
Figure 1: Relationship between selected Omics types
Image source: https://doi.org/10.1016/B978-0-323-54835-9.00015-6
What are Omics data typically used for?
As Omics are rich data that comprehensively capture information at the molecular level (e.g., the protein level) they provide detailed insights into molecular biology and aberrations. In addition, combining multiple Omics data types (multi-Omics or integromics) makes it possible to study the crosstalk between molecular levels. For example, how does DNA methylation (epigenomics) regulate the expression of specific genes (transcriptomics) in a specific tissue type in diseased versus healthy individuals?
In the pharmaceutical industry, traditionally, Omics profiling (generating Omics data from solid tissue, blood, or other sources) is used widely in early research. Here, two examples are presented. The first example is generating transcriptomics data from various tissue samples (e.g., liver, kidney, heart, …) from laboratory animals exposed to a new investigational drug. These transcriptomics data permit study of tissue-specific drug effects: i.e., are genes or entire biological pathways activated by drug exposure as one would expect? The second example is to analyse data from an Omics atlas. A well-known example is TCGA (The Cancer Genome Atlas), which includes Omics data generated from profiling a wide range of tumour types (e.g., colon, breast, prostate, …). A tumour type in TCGA includes many hundreds of patients. A massive resource like this, which can be used to characterize a cancer in detail and understand the frequency of disease-causing molecular events that may be druggable (i.e., new target discovery), is extremely important for researchers working in drug discovery.
Although Omics data are primarily a research tool, these data may become important for regulatory drug approval in the future. For example, the HMA/EMA subgroup report on Bioanalytical Omics states: “there is significant development and potential that these technologies [for proteomics, metabolomics and lipidomics] will be applied also with (hypothesis driven) confirmatory purpose for the development and application of medicines.”. In addition, the FDA developed a cloud-based community analytics platform for sequencing-based Omics data called precisionFDA. This project is expected to generate knowledge to inform future regulatory pathways and decision making.
Omics data in clinical trials
Omics profiling is now routinely performed as part of late-stage clinical trials (i.e., phase 2 or 3). This blogpost discusses the opportunities of having Omics data in clinical trials. It particularly focuses on data types that are modifiable (i.e., that change based on exposure or treatment such as transcriptomics or proteomics) rather than germline genomics (your DNA in every cell), which does not change over time. Others have written about the opportunities of genomics data in clinical trials; and this example is from the data42 initiative.
A clinical trial is used to assess efficacy (does it work?) and safety (is it safe?) of a drug. When complemented with Omics data, measured at multiple time points (e.g., baseline and end-of-trial), these combined data provide a very rich resource for further research.
1. Omics findings as clues for new research
When a clinical trial includes Omics data, one can study the effect of the drug on a very large number of single Omics biomarkers and define the subset of biomarkers that are either repressed (down-regulated) or activated (up-regulated) by treatment. This is important for scientists who are studying mode-of-action of the drug. These data are further valuable because it concerns human data and most early research on mode-of-action is based on animal data.
The study design of a clinical trial is also a major advantage as, compared to real-word data or biobanks, patients in placebo-controlled trials are randomized to treatment, which allows establishing causal conclusions. Also, a trial usually has a longitudinal nature and Omics data can thus be collected at multiple timepoints (e.g., baseline and end-of-trial). When such data are available, one attractive type of analysis is a high-dimensional mediation analysis, which is aimed at finding specific changes in Omics profiles that mediate (i.e., explain) the effect of treatment on a clinical endpoint.
Besides studying effects on single biomarkers, Omics data can be used to investigate known biological (disease) pathways. Information on what single biomarkers (genes, proteins, or metabolites) belong to a specific pathway can be found in bioinformatics databases such as the MSigDB (Molecular Signatures Database). By aggregating biomarkers into known sets as described in such a database, a pathway can be tested in its totality, i.e., is the pathway (de)-activated by treatment as expected? A typical biological pathway can be represented by a few dozen up to two hundred genes. The choice of target tissue used for Omics profiling (e.g., solid tissue, serum, urine, …) is of course very important; and the wrong tissue undoubtfully leads to spurious findings. For example, biomarkers of brain health may not present in urine.
Omics data generated as part of a clinical trial may also point to specific mechanisms not modified by the drug and as such may help define new therapies or combination therapies. Put simply, let’s say drug A modifies two of three key disease pathways but does not act on the third one. This would then suggest that a combination of the drug with one that specifically acts on the third pathway would be superior.
Finally, Omics data may point to patient subsets in clinical trials that respond differently to treatment. This can be achieved using either unsupervised (only using the Omics data) or supervised methods (e.g., by identifying treatment-biomarker interactions in relation to a clinical endpoint). As such, even in a failed clinical trial with no demonstrated clinical benefit of treatment on the ‘average’ trial population, Omics data may help detect the ‘right’ patient subset that benefits from treatment. This is an example of precision medicine. These analytical results can then be used to feedback to Research (reverse translation) and ultimately help initiate a new clinical development project to confirm benefit of treatment in the new target population, which can be identified using biomarkers.
2. Omics profiles as efficacy and safety endpoints
Profiles (models or signatures) derived from Omics data can act as surrogates for hard clinical endpoints in research. There are two potential advantages to having complex Omics signatures as surrogate endpoints. First, a complex signature represents a more sophisticated alternative to a hard clinical endpoint that is not based on a single event or measurement but instead combines many single biomarkers that have specific roles in underlying disease mechanisms. Second, complex signatures may have practical advantages. Changes in Omics profiles may be detectable much earlier compared to the hard clinical endpoints. For example, myocardial infarction (MI) may take years (or even decades) to develop, but changes in a MI risk score based on proteomics may be detectable before the actual event. In addition, hard clinical endpoints may be difficult to capture for example because these require tissue from a biopsy (e.g., liver). A circulating biomarker test would be a more convenient alternative.
Taken together, Omics profiles as surrogate endpoints are an important tool in research especially when the actual clinical endpoint is difficult to assess. Of course, use of these surrogate endpoints beyond research (i.e., as part of clinical drug development) would first require that these biomarkers have been rigorously validated.
3. Omics for data-driven drug repurposing
Drug repurposing (or repositioning) is an important activity in pharmaceutical development. It is the process of finding new uses for already approved drugs. Omics data are an attractive component of a drug repurposing strategy as it may hold important signals not available in other — less rich — data sources.
Practically, this can be done using signature matching where first a molecular signature (fingerprint) of the drug is developed using clinical trial data. Such a signature can be derived from comparing Omics tissue profiles before and after treatment. Second, the molecular signature is compared against signatures of disease risk, which are derived by comparing tissue Omics profiles in diseased versus healthy subjects. Using this approach, one aims to find a high percentage of overlap showing that the same profiles that are up-regulated by treatment are down-regulated in disease, or vice versa. This would then be first suggestive evidence that the drug could be a candidate to treat the disease.
For more on Omics signature matching for drug repurposing, please see this review article.
Summary
In conclusion, Omics data are rich biological data that comprehensively capture molecular mechanisms and disease pathways. Omics profiling is increasingly being used in late-stage clinical trials for drug development to support new scientific research. These Omics data have potential to (1) better understand drug mode-of-action and identify precision medicine opportunities, (2) develop and validate surrogate endpoints and, (3) design data-driven drug repurposing strategies.
