Two DNA tests, two results: where the differences come from

What to expect in this article

1. What happens between your saliva sample and the finished page
2. The four points where nothing is specified
3. Which locations in the genetic material are measured at all
4. Where the meaning is open, the interpretation is open too
5. Who your value is compared against
6. Why the same data produce two numbers
7. How gaps in the measurement data are filled
8. What two reports nevertheless have in common
9. How mybody®x handles these four points
10. What to do when you have two reports
11. Limitations: what this article does not explain
12. What matters when comparing two reports
Frequently asked questions
Sources

What happens between your saliva sample and the finished page

Your DNA is the same in both cases. It does not change when you submit two samples, nor does it change between two analyses. Nevertheless, the report is ultimately not an image of your DNA, but the result of a chain of steps.

This chain has four links. DNA is extracted from the saliva. A chip reads which variant is present at specified positions. A computational step fills in positions not covered by the chip. Finally, the result is assessed against a reference group and translated into text.

Each of these links contains a decision. It is made before your sample arrives at the laboratory, and no binding standard dictates what the decision must be. That is exactly where the differences discussed in this article arise.

Why the reading itself is rarely the problem

Reading a specified position is a technical process with a clearly defined target result. As a rule, there is no disagreement at this point, because two laboratories reading the same position arrive at the same letter.

The difference arises later. It arises where read letters are turned into a statement about you, and this translation is not standardized anywhere. How the reading works technically and how it differs from full sequencing is explained in the article Genotyping or sequencing: the difference.

The difference between a measurement and a statement

A measurement is a finding: at position X, you have this variant. A statement is something else: this variant means this or that for you. Several decisions lie between the two.

When people place two reports side by side, they usually compare the statements. They do not see the underlying measurements at all. That is precisely why a difference appears greater than it is at the raw-data level.

The four points where nothing is defined

Two analyses of the same sample can differ because three things are not defined: which locations in the genome are measured at all, which comparison group serves as the reference, and which formula is used to derive a value from the two. A 2014 paper in Genetics in Medicine attributed differences between providers precisely to these three reasons (Kalf et al., 2014).

There is also a fourth point that this study does not address: filling in missing positions. It comes from a later study and is broken down in Chapter 7 because it introduces a limitation of its own.

The overview does not compare providers. It describes what remains open at each point and how this affects the wording in the report.

The 2014 study is twelve years old. In researching this article, no more recent study could be found that systematically compares reports from multiple providers for the same person. The age of the evidence is stated here rather than concealed.

Which positions in the genome are measured at all

A chip does not read the entire genome. It reads predetermined individual positions at which people frequently differ. The chip's design determines which positions these are.

According to an analysis of consumer genotyping data in the Computational and Structural Biotechnology Journal, most chips for the direct-to-consumer market cover 600,000 to 800,000 positions (Lu et al., 2021). This order of magnitude amounts to less than one percent of the more than 100 million confirmed human variants recorded in the dbSNP reference database.

Why this figure is easily misunderstood

The benchmark matters. The one percent refers to the number of known variants in a database, not to a share of your genetic material. Equating the two turns a sensible selection into a deficiency.

The selection is the real point. A chip specifically measures positions for which there is any research evidence at all. Measuring everything else would increase the amount of data without improving the conclusions.

Two lists that overlap only partially

The same analysis shows how different these lists can be. Two widely used chips overlap in only about 20 percent of the positions measured (Lu et al., 2021). The number and location of the positions read are therefore not the same thing, even if two reports look identical.

The gap can be even larger. The same analysis describes a case in which two datasets with 500,000 and one million positions read had only 100,000 positions in common (Lu et al., 2021). Anyone placing two such reports side by side is comparing largely different questions.

For you as a reader, this has an immediate consequence. If a trait appears in one report but not the other, that does not necessarily mean there is a contradiction. It may mean that the corresponding position was not read in the second case at all.

The selection of positions is therefore one of several reasons for discrepancies. It is neither the only reason nor automatically the most important one.

Where the meaning is open, the interpretation is open too

After the measurement comes interpretation, and that is where we enter an area in which the research itself is not yet complete. A professional society made this clear in an interview in 2025.

In 2025, the German Nutrition Society stated that the function of many gene locations examined in such tests is not fully understood. Where the significance of a location is unclear, its interpretation is also open, and two interpretations can differ without either having been measured incorrectly.

Why many individual locations contribute little

The same interview contains a figure that makes the connection tangible. The approximately 1,000 genetic locations identified to date each have only a small effect on body weight (German Nutrition Society, 2025).

Many small contributions can be summarized in more than one way. Someone who weights twenty locations arrives at a different picture than someone who weights fifty, and both work with the same letters.

What follows from this—and what does not

This does not mean that an analysis is arbitrary. An open research question is different from arbitrariness: it describes room for interpretation with boundaries, not an empty space.

This does not mean, however, that a difference between two reports, by itself, proves anyone wrong. It shows that two different decisions were made at an unresolved point.

What your value is compared against

No value stands on its own. It always relates to a group, and that group is a choice. Which population average serves as the starting point is one of the three reasons identified in 2014 in Genetics in Medicine (Kalf et al., 2014).

You know the principle from everyday life. A height of 1.80 meters is above average or unremarkable, depending on whom you compare yourself with. The measured value remains identical in both cases.

Why the ancestry group matters here

Genetic variants occur at different frequencies in different population groups. A study of 23 chips in six ancestry groups found that a chip tailored to the respective group produced better results than a larger chip that was not (Nguyen et al., Scientific Reports, 2022).

Size alone is therefore not a quality criterion. Fit matters more than scope, and fit is a decision made before measurement.

What this means for your report

If one report says “above average” and the other says “average,” the difference may depend solely on the reference group. Both statements would then be correct because they answer different questions.

The four variables at a glance

Three of these four points are determined before your sample reaches the laboratory. Only the fourth arises during the analysis itself.

Why the same data produce two numbers

The third variable is the least visible. It lies in the calculation rule used to turn many individual positions into one value, and this rule is not included in any report.

The 2014 study found that the providers compared used different formulas and that this affected the predicted values (Kalf et al., 2014). The measured markers themselves were the same.

Two calculation methods usually differ in two respects. One is the number of positions included at all. The other is the weight assigned to an individual position. Neither detail is often included in a report.

Here is an analogy: Two teachers grade the same exam using the same point scale, but weight the questions differently. Both grades are logically justified, yet they are different.

Orders of magnitude from the scientific literature

Sources: Lu et al., Computational and Structural Biotechnology Journal, 2021 · Nguyen et al., Scientific Reports, 2022 · German Nutrition Society, 2025

How to recognize a disclosed calculation

A report does not have to print its formula to be understandable. It already helps if it states how many positions contributed to a result and what the classification is based on.

If both pieces of information are missing, the difference from a second report cannot be explained. All that remains is the observation that two numbers are different.

How gaps in the measurement data are filled

No chip covers all the positions that could be useful for analysis. Therefore, some of the missing positions are supplemented computationally. This process is called imputation—the estimation of unknown positions from neighboring positions that were actually measured.

A study in Genome Medicine examined how much a result shifts as a result. In most cases, the discrepancy remains below five percentile ranks and does not change the interpretation; only in about one percent of cases is it considerably larger (Chen et al., 2020).

Why This Number Is Often Read the Wrong Way

The exceptional case is the more interesting half of the sentence, which is precisely why it is often turned into the main message. The source says the opposite: the usual case is a discrepancy that does not appear in the report.

When comparing two reports, this means that imputation is rarely the explanation. If two analyses differ substantially, the other three variables are more likely to be responsible.

What You Can Take Away in Practice

A report that presents imputed positions as measured conceals an intermediate step. Asking which details were read and which were calculated is therefore a reasonable question to ask of any analysis.

What Two Reports Still Have in Common

Four chapters about differences might leave the impression that everything is a matter of negotiation. That is not the case.

At the level of the letters, there is little room for debate. Your variants at a specific position are a finding, not a matter of discretion, and with proper processing they will be the same in both analyses.

The Part That Remains Stable

The inherited traits that such a test reads remain constant. They are the stable part of the calculation, which is why the reference point of an analysis does not shift on its own. Why this stability means that a DNA test only needs to be taken once, and where the limits of this statement lie, is explained in the article Taking a DNA test only once: what lies behind it.

What changes is the interpretation. This very flexibility explains why two analyses can produce different results at the same time and why the same sample might later be interpreted differently.

Where Agreement Is Expected

For well-studied individual traits, discrepancies are unlikely. The more clearly a site’s function is established, the less room there is for two different interpretations.

Composite statements are different. As soon as many sites with a small contribution are combined into a value, every decision in the calculation process affects the result.

How mybody®x handles these four points

This sentence does not imply an assessment of other analyses. It describes a situation in which every analysis, including its own, is situated. What can differ is how it deals with that situation.

mybody®x (MYBODY Lab GmbH) discloses the procedure. It uses SNP genotyping at more than 700,000 predefined individual positions, not sequencing of the entire genome. This is a statement about the first variable, and it can be verified.

Only part of these identified positions is analyzed, and that part is also quantified. The NutriCare | INFINITY DNA test analyzes 140 genetic variants from a saliva sample and costs €269.00 (as of 08/31/2026; subject to change); the product page lists the reports generated from it.

What the test explicitly does not do

The DNA analysis is intended for nutritional and lifestyle counseling. It is not a diagnostic procedure, does not predict disease, and does not replace a medical examination or consultation.

The scope also has limits that should be stated. The report contains no analysis of protein, dietary protein, or macronutrients, and the tested tolerability reference is limited to four reports in the chapter on metabolic type: alcohol, caffeine, lactose, and gluten.

There is a clear position on the question of interpretation. We do not sell a weight-loss recommendation, but a classification. The test describes predispositions and does not promise success.

The gap that remains here too

There remains a gap concerning the other two variables, and it should be named. Which comparison group underlies a classification and how it is derived is rarely broken down in an evaluation report.

This applies to our report as well as to others. A text about comparability that leaves out precisely this gap in its own case would not do justice to its subject.

What you do when you have two reports

The first instinct is to ask which of the two is correct. This question rarely gets you further because it demands a decision that cannot be made from the material.

The more fundamental question is a different one: Do both reports describe the same thing? The answer is often no as soon as you place the information side by side.

Four questions for both reports

First, check whether the same position is meant. If both reports contain an identifier for the site examined, comparison is possible in the first place; if one lacks it, you are comparing two headings.

Second, check the reference point. A word like “elevated” requires an indication of which group it applies to; otherwise it remains a claim without a benchmark.

Third, check how many positions contributed to the statement. A statement based on five sites and one based on fifty are not the same kind of statement, even if both sound the same.

Lastly, check whether the report distinguishes between measured and supplemented positions. If it specifies this, it makes the fourth lever visible.

Write down the four answers next to both reports before evaluating them. Afterward, you will usually see immediately whether both answer the same question.

When a report belongs in a medical practice

A nutritional DNA report is not a finding about a disease, nor does it answer a medical question. If symptoms led you to this topic, they should be evaluated at a medical practice, regardless of what a report says.

Take both reports with you to such a conversation if you have them. A difference between two analyses is a good reason to ask a specific question and a poor reason to draw your own conclusion.

Limitations: what this article does not explain

The evidence on this question is thinner than the length of the article might suggest. Apart from the source from the German Nutrition Society, it consists of English-language academic publications. No German institution describing the mechanism as a whole could be found.

The main supporting evidence is also old. The study that identifies three of the four levers dates from 2014, and no newer study addressing the same question could be found. This does not weaken the statement; it limits its scope.

There is no reliable figure for the frequency. How often two analyses of the same sample actually differ has not been quantified in the sources reviewed. That is why this article gives no figure for it.

And one limitation concerning this article: This text explains how differences arise. It does not explain which of two specific reports is more accurate in your case. That can only be determined from the reports themselves.

What matters when comparing two reports

If you take away one action, let it be this: Look in both reports for the identifier of the position examined and the indication of which group it was compared with. Two details per statement; that is all you need to start.

The reason is practical. Of the four levers, these are the two that a report can state without revealing its calculations. If you find neither in either report, the comparison cannot be resolved—and that, too, is a result.

The starting point was the question of how two reports about the same DNA can differ. The answer is unremarkable: They do not differ in the measurement itself, but in the decisions made afterward. Once you know these decisions, you read both reports differently.

Frequently asked questions

Sources

1. Kalf RRJ, Mihaescu R, Kundu S, de Knijff P, Green RC, Janssens ACJW: Variations in predicted risks in personal genome testing for common complex diseases. Genetics in Medicine 16(1), 2014, DOI 10.1038/gim.2013.80 – doi.org
2. German Nutrition Society: Personalized nutrition – how does it work? Interview with Prof. Dr. Christina Holzapfel, 2025, accessed on 31 August 2026 – dge.de
3. Lu C, Greshake Tzovaras B, Gough J: A survey of direct-to-consumer genotype data, and quality control tool (GenomePrep) for research. Computational and Structural Biotechnology Journal 19, 2021, DOI 10.1016/j.csbj.2021.06.040 – doi.org
4. Chen S-F, Dias R, Evans D, Salfati EL, Liu S, Wineinger NE, Torkamani A: Genotype imputation and variability in polygenic risk score estimation. Genome Medicine 12, 2020, DOI 10.1186/s13073-020-00801-x – doi.org

The three reasons for differing predictions between providers are taken from [1]. The verbatim quotation about the function of genetic loci and the figure of approximately 1,000 genetic loci are taken from [2]. The information on the number of measured positions, the ratio to the dbSNP variant database, and the overlap between two widely used chips is taken from [3]. The information on imputation and its limitation is taken from [4]. The comparison of 23 chips across six ancestry groups is attributed in the text to the authors, journal, and year: Nguyen et al., Scientific Reports 12, 2022, DOI 10.1038/s41598-022-22215-y. Information on price, scope, and sample type comes from the mybody®x product page, accessed on 31 August 2026. All sources were accessed and reviewed on 31 August 2026. Providers and manufacturers are named in works [1] and [3]; their names are described neutrally here, while the figures are reproduced unchanged.

mybody®x Editorial & Specialist Team

Nutrigenetics Laboratory diagnostics Nutritional science Analysis of genetic data

This article was created by the mybody®x editorial and specialist team. The team brings together nutrigenetics, laboratory diagnostics, and nutritional science. Those who contribute to it are listed on the authors page.

Published on 31 August 2026 · Last updated on 31 August 2026

The DNA analysis is intended for nutritional and lifestyle counseling. It is not a diagnostic procedure, does not predict diseases, and does not replace a medical examination or consultation. Genetic variants describe probabilities in population groups, not fixed outcomes for individuals.