# Why Understand Your Own Data?

> What changes when you understand your own genomic and transcriptomic data, and where the limits are.

The goal of learning to analyze data yourself is to **understand how a result was produced, ask better questions, and revisit it when necessary**.

Here, analyzing it yourself does not mean typing every command by hand. Execution can be delegated to an agent, but the analysis question, the comparison, and the judgment of the result cannot.

[Sid's case](/en/start/case-sid/) illustrates this distinction. Data helped identify possible treatment candidates, and each signal was reviewed alongside other evidence.

## A report summarizes answers, while raw data preserves possibilities

A test report usually contains findings that fit the purpose of the test and can be explained with current evidence. It is not a complete display of every signal. Results with unclear meaning may be marked separately or left outside the reporting scope.

With raw data, you can revisit an analysis when knowledge or methods improve. What can be reanalyzed depends on what kind of data you received. A result table and the original measurements do not preserve the same possibilities.

Raw data does not reveal answers automatically. It also contains signals of uncertain meaning and technical errors. Direct analysis is not a search for a hidden correct answer. It is a process of **separating what is trustworthy from what still needs to be checked**.

## Data suggests candidates, not treatments

Analyzing DNA or RNA from a tumor can reveal altered genes and genes with unusually high activity. These results can help narrow down treatment candidates.

But a clue is not a treatment.

> Find a signal → confirm it another way → review existing research → let specialists assess benefit and risk

A visible target may have no available drug. A drug may exist without evidence that it works in that cancer, or its risks may outweigh its benefit. Treatment decisions combine pathology, imaging, the patient's condition, and treatment history.

The [US National Cancer Institute's guide to biomarker testing](https://www.cancer.gov/about-cancer/treatment/types/biomarker-testing-cancer-treatment) also explains that testing may find no matching treatment, and that a matching treatment may still be ineffective.

Understanding the analysis lets you **ask why one candidate was selected and another was excluded**.

## A value needs a comparison to become meaningful

One number cannot tell you whether it is high or low. Tumor data also needs to be compared with normal tissue or data from other patients before a difference becomes interpretable.

In DNA analysis, comparing tumor cells with normal cells from the same person can help separate inherited variants from changes acquired by the tumor. In RNA analysis, adjacent non-tumor tissue or public resources such as [GTEx](https://gtexportal.org/home/) can provide a reference.

Public data is not always a perfect comparison. Data made by different hospitals or laboratories may include differences in experimental methods. External references are therefore useful for exploring possibilities, while important conclusions require confirmation with other data or tests.

## Your data also contains information about your family

Variants present from birth may also be relevant to parents, siblings, or children. Changes acquired later by a tumor are usually not inherited by family members.

A result that may be relevant to relatives should not be applied to them directly. Each person should receive genetic counseling and confirm the finding with a validated test.

Genomic data cannot be changed after a leak like a password can. When storing it yourself, follow at least these principles:

- Do not upload it to a public Git repository or an online service whose provenance you do not know.
- Encrypt your computer, external storage, and backups.
- Share only the necessary files and check whether names, dates of birth, or other identifiers remain.

The [NHGRI overview of genomic privacy](https://www.genome.gov/about-genomics/policy-issues/Privacy) explains that even de-identified data may be reconnected to a person or family.

## Record enough to check the analysis again

Keeping the analysis code makes it possible to rerun the same process and review it with someone else. This is a major advantage when developers learn to analyze data directly.

Code alone is not enough. Record which data was used, which tools and criteria were applied, and what was excluded along the way. Only then can someone follow how the result was produced.

## Why understanding it yourself is worthwhile

Learning to analyze data directly lets you do three things.

1. **Check**: understand which data and criteria produced a result.
2. **Ask**: question what was included and what was left out.
3. **Revisit**: review the result when new data or knowledge becomes available.

The goal is to **participate more effectively in decisions about your own data**.

Next: [Principles We Follow](/en/start/principles/) explains how we handle real cases, personal data, and AI output.