Omics biomarkers and early diagnosis: when happiness is possible

One of the important goals of omics technologies (including transcriptomics, proteomics, metabolomics) in medicine is the search for biomarkers for the diagnosis of common non-infectious diseases, mainly malignant tumors. They often talk about early diagnosis, that is, about those cases when the disease is detected for the first time and at an early stage. But does it always make sense to use molecular technologies to search for early markers? I want to talk about what parameters characterize molecular biomarkers and their combinations, for what cases they should be developed, and also talk about the decade of development of post-genomic technologies in medicine and about my own experience in cancer proteomics. This material is addressed mainly to biologists and physicists who are involved in omics technologies, but do not always go into medical details. It will help them avoid unnecessary work in terms of clinical problem setting.

One of the main goals of postgenomic, or omics, technologies in medicine is the development of biomarkers for disease diagnosis [1]. Postgenomic technologies usually mean high-throughput measurement of gene expression products—transcriptomics and proteomics [2]. Metabolites are considered indirect products of gene expression, and their analysis - metabolomics - is also often considered a postgenomic method, although, strictly speaking, the metabolites of the human body contain many xenobiotics that are not directly related to the functioning of the genome [3].

“Biomolecule” has already described what omics technologies are and how “big data” has transformed the biology of today and tomorrow: “Omics is the era of big biology” [4].

On the other hand, the question arises of what biomarkers are, although this word is well-known and its meaning is intuitively clear. A biomarker is a measured parameter, the qualitative or quantitative characteristics of which indicate the presence or absence of a disease or condition. In a broad sense, even a black eye can be a biomarker. In clinical medicine, such biomarkers are called symptoms. But in biochemistry there are other, molecular biomarkers, that is, molecules measured in different ways. And these molecules can be very different - from simple ions to high-molecular proteins, for example, antibodies to various antigens [5].

Technologies for identifying various biomolecules in a high-throughput mode are rapidly progressing [2], [6–10]. Computational methods for processing “big data” are being improved and become more complex [11]. However, technological progress does not negate the importance of balancing omics methods with preparatory clinical work.

A significant part of the studies uses patients with diseases (for example, with a malignant tumor) as comparison groups, and practically healthy people or people with chronic pathologies not related to the disease of interest (in our case, cancer) as controls. In other words, with the help of omics technologies, tests are being developed for early diagnosis - in order to distinguish patients from healthy ones as early as possible, to detect the disease at its early stage [12]. But is this comparison always relevant? What properties should a biomarker have in order for it to be useful for early primary diagnosis, that is, when the disease is detected for the first time?

I noticed this problem while working on grant applications for basic medicine [13]. In an attempt to develop new diagnostic tests, their authors have often proposed comparing samples from patients with different stages of malignant tumors with samples from relatively healthy people. At the same time, the methods that the researchers had at their disposal clearly did not correlate well with the idea of ​​screening - a large-scale examination of a large number of healthy people to identify a relatively rare disease. After all, even the most common malignant tumors, as we will see later, remain a relatively rare phenomenon.

Pharmacological properties of the drug Omix

Tamsulosin selectively and competitively blocks postsynaptic α1A-adrenergic receptors located in the smooth muscles of the prostate gland, bladder neck and prostatic urethra. Reduces the tone of the smooth muscles of the prostate gland, bladder neck and prostatic urethra, improving urine outflow. At the same time, the severity of symptoms of obstruction and irritation associated with benign prostatic hypertrophy decreases. The therapeutic effect develops 2 weeks after the start of treatment. Tropism for α1A-adrenergic receptors located in the bladder is 20 times greater than its ability to interact with α1B-adrenergic receptors located in vascular smooth muscles. Due to its high selectivity, it does not cause a clinically significant decrease in systemic blood pressure both in patients with hypertension (arterial hypertension) and in patients with normal initial blood pressure. After oral administration, it is quickly and completely absorbed into the gastrointestinal tract. Bioavailability is approximately 100%. After a single oral dose of 400 mcg, the maximum concentration of the active substance in the blood plasma is achieved after 6 hours. At steady state (5 days after the course of administration), the maximum concentration of the active substance in the blood plasma is 60–70% higher than after a single dose of the drug. The degree of binding to blood plasma proteins is about 99%. Tamsulosin is not subject to the effect of primary passage through the liver and is slowly biotransformed in the liver with the formation of pharmacologically active metabolites that retain high selectivity for α1A-adrenergic receptors. Most of the active substance is found in the blood unchanged. Excreted in the urine, 9% of the dose is excreted unchanged. Tamsulosin with a single dose is 10 hours, final half-life is 22 hours.

What is needed for the test to be suitable for early diagnosis?

Without repeating the textbooks, let me remind the reader of the commonly used diagnostic test parameters—accuracy, sensitivity, and specificity (see box). Another important parameter characterizing the new test is the diagnostic value of a positive result (DPV, positive predictive value) [14]. It is the ratio of the number of true positives (when the disease is found correctly) to the sum of all positives, true and false:

where DCPR is the diagnostic value of a positive result, IP is the number of true positive results, LP is the number of false positive diagnostic results.

DTsPR helps to assess the feasibility of using the test for a general examination of any group of people, that is, screening. Its low values ​​can make the diagnostic method costly and even unsafe from the point of view of the overall benefit for the subjects. But we’ll dwell on this below.

Biomarker diagnostic parameters

A newly developed diagnostic method must be superior to all existing ones, and for this it must be compared with the “gold standard” - the best method that gives an (almost) error-free answer (Fig. 1). However, if there is a gold standard, then why a new test? And then, for example, that it may be more clinically beneficial or allow you to solve the problem faster, easier and cheaper. Thus, if the gold standard is an invasive study during surgery or generally post mortem, its diagnostic significance, alas, is low.


Figure 1. How to evaluate the performance of a diagnostic test? What if it worked incorrectly, calling a healthy person sick or vice versa? To do this, we need to know what is really happening. A test that is very reliable - the “gold standard” - will help with this. By comparing new results with this standard, we can calculate errors and decide whether our diagnostic method has a future.

illustration by Alena Belyakova

The main characteristics of the new test are its diagnostic accuracy, sensitivity and specificity. Accuracy is the proportion of correctly diagnosed cases among all those examined, both sick and healthy. Sensitivity is the proportion of correctly diagnosed sick people, and specificity is the proportion of healthy people.

One might guess that sensitivity and specificity, as opposed to accuracy, depend on the cutoff value of the biomarker or combination of markers. A threshold value is, for example, the concentration of a molecule in the blood at which (or higher values) we consider a person to be sick. In the case of a combination of markers, such a threshold will be a more complex function that takes into account the values ​​of several parameters. If we greatly reduce the threshold for measuring the biomarker concentration, we will detect all patients (sensitivity 100%), but we will misdiagnose more subjects who do not suffer from the disease. Thus, when describing biomarkers, it makes no sense to present their sensitivity and specificity separately.

Results of an error-free examination (“gold standard”)
Total number of examinedGold standard +Gold standard -
New test resultsNew test +True positive result (IP)Type 1 error - false positive result (FP)
New test -Type 2 error - false negative result (FN)True negative result (TR)
Characteristics of the new testAccuracy* = (IP + IR) / Total number of examined Sensitivity = PI / (IP + LO) Specificity = IR / (IO + LP)
* It is important not to confuse diagnostic parameters with analytical ones. For example, analytical accuracy is a function of instrumental measurement error. The analytical sensitivity of a method is the detection limit of the compounds of interest by this method, and is not directly related to diagnostics.

Let's imagine a situation where a hypothetical omics test (for brevity, this is a test developed using genomic and post-genomic technologies) showed outstanding results and distinguishes patients from apparently healthy people with sensitivity and specificity of 98%. Let's assume that we are talking about a common malignant tumor with a population incidence of 100 cases per 100,000 population: these numbers are valid, for example, for prostate cancer in the countries leading in its incidence [15]. Let's screen these 100,000 people. With a sensitivity of 98%, out of 100 truly ill patients, 98 will be correctly diagnosed and only two will be missed. With a specificity of 98% of the 99,900 healthy people we examine, 2%, that is, 1,998 of those examined, will be incorrectly diagnosed with cancer. It is easy to calculate that the DCPR in this case will be only 4.6%, and almost 2,000 people will suffer moral damage and will be assigned an expensive and possibly traumatic examination (Fig. 2). Note that 98% sensitivity and specificity are parameters that are usually not achievable in studies.


Figure 2. Even with a diagnostic specificity of 98%, the number of false-positive diagnoses of relatively rare diseases will be significant in terms of the absolute number of subjects examined. Such people, who do not actually have the disease, will undergo an expensive and traumatic examination.

illustration by Alena Belyakova

The conclusion suggests itself: for an omics test to be suitable for early diagnosis of a disease, several conditions must be met (Fig. 3). Firstly, in the case of 100% specificity, the DCPR will automatically reach 100% (see formula above). This is possible in the case when a biomarker or a certain combination of biomarkers is detected exclusively in pathology, but not normally, that is, the qualitative principle “is there or not” applies. An example is somatic driver mutations that occur in the genome of a malignant tumor [16]. The probability of finding them in the germinal genome of an adult who does not have a malignant tumor is vanishingly small. If omics tests involve detecting cancer by detecting mutant DNA, RNA and protein sequences in any biological material, these tests have a chance of being used for early diagnosis.


Figure 3. Selecting a clinical trial strategy for developing an omics test. Conditions that must be met for the test to be suitable for early diagnosis are presented. Clinical needs guide study design: different study cohorts should be selected depending on the task.

illustration by Alena Belyakova

Another example of substances, the detection of which claims to have 100% diagnostic specificity, are oncometabolites that are synthesized exclusively in tumor cells due to mutations that modify the activity of enzymes [17]. Do not forget that with absolute specificity, the test must maintain acceptable diagnostic sensitivity. It should be noted that to date, tests based primarily on mutant tumor DNA have reached the point of being put into practice. They can also analyze genomic regions with base hypermethylation characteristic of tumors [18]. With this phenomenon, the transcription of tumor suppressor genes is “cancelled.”

For example, with some reservations, the Cologuard stool test approved for use in the United States can be considered a screening test for colorectal cancer [19]. This test is multiplex and also combines the use of various methods for detecting different molecular events. It involves quantitative measurement of DNA with a mutation in the KRAS gene, assessment of abnormal methylation of the NDRG4, BMP3 and β-actin genes, as well as an immunoassay for hemoglobin.

If the test's specificity is not absolute, it can still be used to screen for conditions more common than cancer. However, high incidence rates are characteristic of infections, the principles of diagnosis of which are usually well defined and differ from omics.

The clinic uses a number of biomarkers to diagnose malignant tumors, which are sometimes positioned as screening ones. In this capacity were, for example, various forms of prostate-specific antigen (PSA), measured in blood plasma, and other “cancer antigens” (CA), discovered in the 80s of the last century by immunological methods. However, the diagnostic performance of these molecules is such that they would not pass regulatory approval today for most of the indications for which they were previously used. Therefore, medical communities in the vast majority of cases no longer resort to evaluating “old” tumor markers for early diagnosis. For example, PSA testing is no longer recommended for widespread prostate cancer screening [20]. Imaging methods [21] and other approaches, including the physician’s determination of combinations of symptoms, play a leading role in the early diagnosis of many tumors.

Special instructions for the use of the drug Omix

Use with caution if there is a tendency to orthostatic hypotension, severe liver dysfunction and creatinine clearance ≤10 ml/min. If signs of orthostatic hypotension (dizziness, weakness) appear, it is recommended to sit or lay the patient down. Before starting treatment, it is necessary to verify the diagnosis. Caution should be exercised when driving vehicles and performing work that requires high concentration (due to the possible development of dizziness).

Omics tests: from early diagnosis to risk assessment

In most cases, panels of post-genomic biomarkers provide results that do not allow their use for early diagnosis of diseases. However, there is a large area in medicine that is in dire need of the development and use of biomarkers. We are talking about prognostic and predictive biomarkers that acquire value after diagnosis [22]. Without such markers, it is difficult to imagine personalized medicine, which involves dividing patients with the same diagnosis into groups and rationally choosing different types of prevention and treatment for them [23], [24]. Despite the semantic similarity of the terms mentioned, they serve to distinguish between biomarkers that predict the severity of the disease and the likelihood of relapse (prognostic), and biomarkers that predict response to therapy (predictive).

The history of the first proteomic biomarkers is indicative. In the early 2000s, it was proposed to use mass spectra of almost unprocessed plasma obtained using direct MALDI-TOF mass spectrometry to classify blood plasma samples [25]. For example, in a study in which I took part back in 2005, the mass spectra of samples from healthy people and from patients with ovarian cancer were compared [26]. It soon became clear that although this test was capable of recognizing the disease, its characteristics were not suitable for early diagnosis. Therefore, tests of this kind are no longer considered screening tests. The approach was rightly criticized for working in a barcode format without identifying the components of the diagnostic profile, and when they were identified, they turned out to be mainly nonspecific inflammatory response proteins with high concentrations in the blood.

One of the articles of our special project “12 biological methods in pictures” - “Proteomics” [2] talks in detail about mass spectrometry. - Ed.

One of the American companies, which owned patents for some components of mass spectrometric profiles of blood plasma in ovarian cancer, shifted the focus of development towards a predictive test. Findings made using MALDI-TOF profiles formed the basis of the approved OVA1 test, which measures levels of inflammatory proteins and the “old” marker CA-125 [27]. Such a test is in no way considered as a screening test, but rather predicts the risk of tumor malignancy in patients with previously identified (for example, during ultrasound examination) formations in the pelvis. Of course, there is no talk of separating sick and healthy people using this test. OVA1 helps determine the extent of surgery and treatment strategy. This test has not become indispensable on the market due to the discovery of a competitive biomarker, the HE4 protein, using biochemical rather than proteomic methods [28]. In combination with CA-125, it performs functions similar to the OVA1 test as part of a duplex test with the original ROMA risk assessment algorithm [29]. This test is also available to Russian patients.

Biomarkers: terms

Biomarker is a natural characteristic, including a molecule or gene, that can be used to determine a specific pathological or normal but altered state of the body.
A diagnostic biomarker is used to establish a diagnosis, that is, to detect a disease or accurately determine its separate, independent (nosological) form. A prognostic biomarker is used to predict the development of a disease in the future or the nature of the course of an already developed disease. A predictive biomarker is used to predict a patient's response to a specific treatment. Screening in medicine is the examination, using a diagnostic test or method, of a large number of people who do not have signs of disease. Screening is needed to detect the disease as early as possible, in an asymptomatic form. Well-known examples of tuberculosis screening are the tuberculin test (the notorious Mantoux test) and chest fluorography. The number of projects being carried out in our country to create diagnostic omics tests, as appears from applications submitted to Russian scientific foundations, is growing. They are often managed by specialists responsible for complex technology, that is, physicists, chemists or biologists, assisted by doctors. Unfortunately, in a number of such projects, despite the enormous efforts associated with the technical part of the work, modern instruments and reagents, the clinical task is incorrectly and superficially formulated, which may turn out to be obviously impossible. For example, to create a diagnostic omics test, those tumors are selected whose early diagnosis is already achievable at a sufficient level using imaging methods.

In this article, where I mention some episodes from the development of the scientific direction in which I participated [26], [30], [31], I would like to encourage the community involved in post-genomic technologies to effectively contact enlightened clinicians to set the right, modern medical goals. The diagnostic parameters of omics analyzes discussed in this work in relation to malignant tumors apply to any non-infectious diseases for which the diagnosis of molecular tests makes sense. In many cases, adequately defined tasks of omics technologies, especially proteomics and metabolomics, are associated with dividing patients with an already diagnosed diagnosis into groups for more correct disease management, risk assessment and the prescription of effective therapy.

Rating
( 1 rating, average 4 out of 5 )
Did you like the article? Share with friends:
For any suggestions regarding the site: [email protected]
Для любых предложений по сайту: [email protected]