The most common question I have seen from people is “what can I do with my DNA?” As dimenwarper and Bryan Johnson put it on X:

…and for the most part, this is true. The questions we can ask are mainly about the genome structure itself, such as:

• What genes do I have?
• Do I have some letters that are bad?
• What “mutations” do I have relative to population averages?
• What drugs do I have bad reactions to?
• What will my kids inherit?

But there is a lot more we can do! And the things we can do will expand as the costs collapse through wet lab automation, personal cell storage, and live editing + programming of your own biology. Just like the internet, crypto, AI and other information-based advancements, the genome (and molecular biology at large) will enable all kinds of new technologies, products, and services.

• I. Finding disease associated variants
• Case 1
• Case 2
• II. Predict drug metabolism
• Case 1
• Case 2
• III. Identify biological pathways to investigate
• Case 1
• Case 2
• IV. Prioritize molecular targets
• Case 1
• V. Detect viral DNA that has integrated into your genome
• Extensions
• The Pangenome Graph
• References
• Research papers and clinical resources
• Biology databases and tools
• Pangenome resources
• Other links cited in the text

I. Finding disease associated variants

This is the most obvious one. Humans share 99.9% of the same DNA sequence. The remaining differences, albeit 0.1%, are enough to contain millions of genetic variants that very few others have. In addition, these 0.1% differences can also combine in specific ways that can cause disease. A core objective of whole-genome sequencing is to identify these and understand their mechanics.

There are two categories of genes- common genetic variants and rare pathogenic variants. Common genetic variants have small individual effects but can collectively influence your probability of developing a disease. These are usually analyzed using polygenic risk scores, which are probabilities assigned to these common genetic variants about what ‘outcome’ they result in in your biology. Rare pathogenic variants have much larger effects, like disrupting a single protein enough to cause an inherited disease.

The second category is obviously the one we care about the most. To figure this out, we can use the central dogma of molecular biology! DNA → RNA → proteins. We can sequence someones genome (DNA), then sequence the expressions of those genes (RNA), measure the misfolded proteins from those RNA’s (proteomics), investigate what happens when the resulting protein doesn’t function correctly, simulate the pathways that are now disregulated, and finally model what that means for your whole body. The body is shockingly sensitive to chemicals (hence why people respond so differently to the same molecule and dosage in medication), so even a misfolded protein can cause serious system-level problems that result in you feeling something or observing something in/on your body.

Case 1

As a real world example of how important even just sequencing is, in September 2026, researchers published a study in Nature Medicine examining 100 critically ill children using rapid WGS. They identified a molecular diagnosis in 53 patients. Five patients had more than one genetic diagnosis. However, one interesting case involved a three-month-old infant experiencing infantile spasm (F32). The researchers, through WGS, were able to identify pathogenic variants affecting two separate genes:

1. SLC19A3: Thiamine metabolism dysfunction syndrome 2

This gene encodes thiamine transporter 2, a protein responsible for transporting vitamin B1 into cells. Thiamine is required for several metabolic processes involved in cellular energy production. Inside the cell, and enzyme converts thiamine into thiamine prophosphate (TPP), which is a cofactor required by enzymes to convert pyruvate into acetyl-coA for cellular energy production.

When the transporter is dysfunctional (misfolded), cells can receive insufficient thiamine to support normal energy metabolism. The brain is especially vulnerable because neurons have high energy requirements.

People with this disorder can develop seizures, spasms, muscle rigidity, and many other movement-related problems that can eventually cause confusion and reduced consciousness. Brain MRI scans have shown symmetrical lesions in the basal ganglia, especially in the caudate nucleus and theputamen, which are structures heavily involved in regulating movement.

In severe infantiale cases, the disorder can produce infantile spasms, developmental regression, lactic acidosis, and brain injury. Untreated episodes can progress to coma or death. The relationship between SLC19A3 dysfunction, impaired thiamine-dependent metabolism, and neurological disease is well established. The precise chain is still being investigated, but this gene is very much related to the infant spasms the infant was having, and the infant happened to have the gene.

2. BTD: Biotinidase deficiency

BTD encodes biotinidase, an enzyme involved in recycling biotin, aka vitamin B7. When this enzyme doesn’t function properly, the body can develop deficiencies that affect metabolic pathways for neurological processing (your nervous system + brain’s computation).

The sequencing revealed both of these. Within 83 hours, they were able to identify what was wrong with the infant and administer the correct solution, which happened to be vitamin B1 and B7.

Case 2

There is another example from the study, an infant we will call F81. This individual had a swollen belly (abdominal distention), calcification of the adrenal glands, and progressive deterioration from the disease progression. This essentially means that they couldn’t properly break down fats, so the fats accumulated and began damaging their organs.

Using WGS, the researchers identified a homozygous exonic deletion in the LIPA gene, thus diagnosing Wolman’s disease. This means that both copies of the gene were affected (homozygous) and part of the protein-coding sequence was missing (exonic deletion).

LIPA encodes lysosomal acid lipase, an enzyme responsible for breaking down cholesteryl esters and triglycerides inside lysosomes. Lysosomes are membrane-bound organelles that contain enzymes responsible for degrading and recycling cellular materials. Usually, the lysosomal acid lipase breaks down these lipids into smaller molecules that the cell can reuse, but when the enzyme is severely deficient, the lipids accumulate inside of the lysosomes instead. So, the fats accumulate, without any recycling happening.

As a consequence, the accumulation starts to damage tissues, especially the liver, spleen, adrenal glands, and gastroinstestinal system. Calcium salts then deposit into those damaged areas. This can be fatal to an infant in <12 months.

To fix it, the clinicians initiated enzyme replacement therapy and changed the infants diet. Specifically, they implemented an enzyme replacement therapy called sebelipase alfa, a recombinant form of human lysosomal acid lipase. Rather than correcting the DNA mutation itself, the enzyme replacement therapy supplies functional copies of the deficient protein.

This is delivered intravenously to be taken up by cells, where it is transported to lysosomes. Once there, the replacement enzyme can break down the lipids that would otherwise accumulate.

The study reports that early treatment stabilized the infant's clinical progression. It also notes that genomic diagnosis allowed treatment to begin approximately two weeks before conventional enzymatic confirmation.

You can search on X for a ton of different cases like this, but WGS is very valuable for disease identification. It will be even more exciting to see work done for upgrading humans, not just fixing problems that are resultant from deformities in the genome.

II. Predict drug metabolism

Another valuable avenue for WGS is predicting how your body will respond to medications. Two people can take the same drug, at the same dose, and have completely different effects. One point that I will constantly reiterate is that the body is extremely sensitive to chemicals. One person might respond like the expected population average, while someone else experiences no benefit, and another experiences severe side effects. Part of these differences in response come from genetics.

When a drug enters the body (blood, stomach, tongue, lungs), enzymes and transporters determine how it is absorbed, activated, distributed, and cleaned up. The genes encoding the proteins at each step contain variants that change how efficiently they function, if at all. Remember, the body is extremely sensitive, chemically.

This field of study is called pharmacogenomics. Unlike polygenic risk scores, which are matching genes to population phenotypes and assigning percentages, pharamcogenomics relationships are simple causal chains. I have this mutation in my genome, which causes this problem, which results in my metabolism doing this other thing, which prevents me from using some drug for some specific reason, meaning we must use this other drug at this specific dose.

Case 1

There is an example that involves the DPYD gene and a chemotherapy drug called fluorouracil (5-FU). This drug is used to treat several cancers such as colorectal, breast, and gastrointestinal. It interferes with the function and production of nucleic acids, preventing cancer cells from dividing normally.

However, like a lot of drugs, it also affects healthy cells. Under normal conditions, the body relies on an enzyme called DPD (dihydropyrimidine dehydrogenase) to break down >80% of the administered fluorouracil. DPD is encoded by the DPYD gene.

(8F5W)

See the connection? DPYD gene → DPD enzyme production → prevents cancer cells from dividing. If this isn’t there, we need to step in with fluorouracil.

Some individuals inherit genetic variants that reduce or eliminate DPD enzyme activity. If they receive a normal amount of fluorouracil, their bodies may be unable to eliminate the drug efficiently. This results in excessive drug exposure and toxicity, which will damage those healthy cells.

So how does a genomic test pick this up? Once we sequence someone, we can look for a specific variant, rs3918290. This mutation affects a splice site, which is a region of DNA that the cell uses when transcribing RNA. The variants disrupts this process by causing exon 14 to be skipped. Because the RNA transcribes incorrectly, we get an abnormal protein that has abnormal enzyme properties. This literally compounds into a systematic problem. You can see the specifics below.

Normal DNA (GRCh38 reference strand):

5'-TGTTTTAGATGTTAAATCACACTTA
   CGTTGTCTGGAAAGTCAGCCTTTAGT-3'

Mutated DNA:

5'-TGTTTTAGATGTTAAATCACACTTA
   TGTTGTCTGGAAAGTCAGCCTTTAGT-3']

Standard drug dosing assumes that the body can metabolize a medication within a predictable range. But for someone carrying this mutation, that assumption can be dangerous. According to the CPIC guidelines, someone with one nonfunctional copy of DPYD may need to start at 50% of the standard fluorouracil dose, while someone with two nonfunctional copies should generally avoid the drug entirely.

In a real world setting, you would go from a single nucleotide in someone's DNA to predicting how their body processes a chemotherapy drug. Then you would use that information to change the treatment before it was administered. That's the value of sequencing your genome: identifying molecular differences that can directly inform what to put in your body, how much, and why.

Case 2

Another example involves Keytruda (pembrolizumab), an antibody-based cancer treatment that blocks a protein called PD-1 on immune cells. Typically, PD-1 helps prevent the immune system from attacking healthy tissue. However, cancer cells can exploit this mechanism to prevent immune cells from attacking them.

Keytruda blocks this interaction, allowing immune cells to attack certain tumors more effectively.

Some of these tumors have defective DNA repair machinery, often involving genes such as MLH1, MSH2, MSH6, and PMS2. When these systems have problems, mutations accumulate throughout the cancer's genome. This then produces a condition called MSI-H (microsatellite instability-high), which is associated with an increased number of abnormal proteins that the immune system can recognize.

We can detect these abnormalities by sequencing the tumor's DNA!

In a 2020 study in the New England Journal of Medicine, patients had colorectal cancer that spread to other parts of the body. Their tumors all had defective DNA repair systems. Patients receiving Keytruda went a median of 16.5 months without their cancer getting worse, compared with 8.2 months for those receiving chemotherapy.

By sequencing, they identified a weakness, then were able to specifically attack it and double the average length of time they can live. This doesn't guarantee the treatment will work, but it does who how we can use someone's molecular data to make better treatment decisions and conquer the disease.

III. Identify biological pathways to investigate

We now know the genome can be used to identify disease-associated variants and predict how our bodies might respond to certain medications. But genes and proteins don't operate independently. They interact with thousands of other molecules through interconnected pathways.

A pathway is a sequence of chemical interactions that produces some biological outcome. For example, when your immune system detects a pathogen, proteins bind to receptors, activate other proteins, and eventually trigger the production of molecules that coordinate an immune response, all through a chemical cascade.

Researchers have built enormous databases documenting these reactions, such as Reactome, KEGG, Gene Ontology, and STRING.

We can take genes identified from our genome, locate the pathways those genes participate in, and investigate how different mutations might affect the larger biological system. The outputs are really the phenotypes we see in someone.

Case 1

In 2024, researchers published a study in Nature Communications analyzing the protein-coding DNA of 350,770 adults from the UK Biobank.

They were investigating immune-mediated diseases, where the immune system starts attacking the body's own tissues. These diseases include rheumatoid arthritis, psoriasis, asthma, and several autoimmune disorders.

Using whole-exome sequencing, which captures the protein-coding portions of DNA (introns), the researchers identified 162 genes significantly associated with 35 different immune-mediated diseases. They wanted to understand how the genes interacted, which biological processes they affected, and whether those processes could reveal potential treatment targets.

The researchers first used a database called STRING, which contains information about interactions between proteins. STRING allowed the researchers to construct a network showing how the disease-associated genes were connected through their encoded proteins.

By using Reactome (molecular reactions and pathways), KEGG (maps biochemical and signaling systems), and Gene Ontology (organizes genes according to their biological functions), the researchers were able to compare their gene networks against these databases to determine whether a particular biological process appeared more frequently than expected.

Case 2

In a 2024 Nature Communications study, researchers analyzed the protein-coding genomes of 350,770 individuals, identifying 162 genes associated with 35 immune-mediated diseases. Rather than examining these genes independently, they used STRING to construct protein interaction networks, then Reactome, KEGG, and Gene Ontology to identify the biological processes represented within those networks. Three major functional groups emerged: immune signaling, urate metabolism, and antigen processing. What initially appeared to be hundreds of separate genetic associations could now be investigated as interconnected biological mechanisms!

The researchers then incorporated single-cell RNA sequencing data from Tabula Sapiens to determine where these genes were being expressed. For example, LRP1, a gene associated with asthma in their analysis, showed expression in neutrophils in blood, while TET2 was expressed across several immune-cell populations. This additional information provided cellular context so that they could connect genetic associations to molecular pathways to identify the cell populations involved in those processes (ironically, this end to end service is what we provide at oak.bio).

We can apply the same methodology to an individual's genomic and transcriptomic data. We can identify variants, investigate their associated proteins and pathways, and determine which biological processes warrant further measurement. Pathway enrichment alone doesn't establish dysfunction or causality, but it allows us to move beyond isolated genetic associations toward testable hypotheses about the molecular systems underlying a phenotype, then design therapeutics or improvements.

Pharma companies then developed monoclonal antibodies that bind to PCSK9 and prevent it from interacting with LDL receptors. One of these drugs, evolocumab (Repatha), was evaluated in the FOURIER clinical trial, involving 27,564 patients with cardiovascular disease. When added to statin therapy, evolocumab reduced LDL cholesterol by approximately 59% compared with placebo and reduced the risk of the trial's primary composite cardiovascular outcome by 15% over a median follow-up of 2.2 years.

So why is this relevant to genetics and justifying the sequencing of your own genome? Well, the therapeutic strategy originated from observing naturally occurring genetic differences in humans. Researchers identified a mutation, connected it to a protein and its biological function, determined that reducing the protein's activity was beneficial, and developed a molecule that reproduced part of that effect.

We can now investigate similar relationships computationally! Given a genome, we can identify variants associated with a phenotype, use pathway databases to understand their molecular consequences, and query Open Targets or ChEMBL to identify proteins with supporting genetic evidence and existing compounds that influence them. If there aren’t any, we can use the generative chemical models like Boltz2 and AlphaFold3 to target them.

To be clear, this doesn't guarantee that a proposed intervention will work in an individual, but it provides a systematic framework for deciding which molecular targets to inspect further. Upstream of this is wet lab testing, clinical trials, and manufacturing of the chemical.

IV. Prioritize molecular targets

In this article, we have identified disease-associated variants, investigated how they influence drug metabolism, predicted their effects on a persons genome, and connected genes to the biological pathways they participate in. The next question is whether we can use this information to identify specific molecules worth targeting with a therapeutic intervention.

A molecular target is typically a protein whose activity we want to increase, decrease, neutralize, or otherwise modify. Researchers use databases such as Open Targets, ChEMBL, and DrugBank to connect genetic evidence with protein function, existing drugs, and experimental measurements. By combining these resources with genomic and transcriptomic data, we can prioritize targets based on their relationship to a disease, the direction in which their activity should change, and whether they can be safely manipulated.

Case 1

In a 2006 study published in the New England Journal of Medicine, researchers investigated rare genetic variants in PCSK9, a gene encoding a protein that regulates how the liver removes LDL cholesterol from the bloodstream. PCSK9 usually promotes the destruction of LDL receptors, reducing the liver's ability to clear cholesterol. However, some individuals inherit mutations that significantly reduce PCSK9 activity, allowing more LDL receptors to remain available.

The researchers found that individuals carrying certain loss-of-function mutations had approximately 28% lower LDL cholesterol and an 88% lower incidence of coronary heart disease over 15 years. Whether this is causation or correlation is up to the statisticians, but it did provide genetic evidence that reducing PCSK9 activity could protect against cardiovascular disease. However, in terms of pharma indications, the molecular target was pretty clear: inhibit the PCSK9 protein to increase LDL receptor availability and remove more cholesterol from circulation.

V. Detect viral DNA that has integrated into your genome

The history of biology is, in part, a war between viruses and cells. Bacteria have been fighting viruses for billions of years, and that same conflict extends to eukaryotic cells like human skin cells, muscle cells, and neurons. This evolutionary war has shaped our DNA, the proteins our cells express based on specific inputs, and the mechanisms we use to defend ourselves against infections.

Some viruses operate by inserting their own genetic code directly into the DNA of the cells they infect. Sometimes these insertions make their way into future generations, becoming a permanent part of our genome. In fact, roughly 8% of human DNA comes from ancient retroviruses. Most of these sequences can no longer produce viruses, but some have been repurposed by evolution. Certain viral genes even help us form the placenta during pregnancy!

With whole-genome sequencing, we can look for these viral sequences in our own DNA, identify where they are located, and investigate whether they influence nearby genes. The interesting question is whether any of this viral DNA is still doing something: changing which proteins we express, influencing our immune system, or contributing to disease.

If you want me to analyze your genome, recommend sequencing companies to use, or anything computational, feel free to message me on X or email me at bradley@oak.bio. I am very responsive and answer every request.

Extensions

The Pangenome Graph

Today we have companies that are trying to build the valuable use case of the whole genome, just like the internet in the early days and crypto in the early 2010s. Information is fundamentally a network problem. When you get more users connected to this network, thus more people adding to this data structure (graph), you start to get emergent properties. Products and use cases you never envisioned (self-driving cars, rockets).

There are academic efforts underway like the pangenome project, which represents a bunch of different genomes as a graph. And this is surprisingly very recent!

As you can see, the sample size for these is underwhelmingly small. I reference a new business model for genomes in this article, but the more genome’s we have, the more use cases emerge.

But what can we do today, with <100M fragmented whole genomes globally?

1. we don’t know yet, but it will be big
2. we can pair DNA with other data, like RNA and proteins, and get a useful understanding of someone’s genes + what genes are being expressed, and at what amount
3. but we have very little action on DNA alone, like websites being added for Google, or suppliers for Amazon, or people from your network posting for Facebook (chicken and egg problem)

Today, there are use cases when you pair it with other layers of data of biology. The genome, alone, however, mainly just tells use the molecules in it- the ones we know that lead to pretty bad diseases down the line, and some that lead to really positive things, like energy levels from hyperactivity.

References

Research papers and clinical resources

• Citywide implementation of a rapid whole-genome sequencing program for critically ill pediatric patients
• DPYD pharmacogenetics and fluoropyrimidine dosing guidance
• Protein structure 8F5W
• Pembrolizumab in microsatellite-instability–high advanced colorectal cancer
• Genetic associations with immune-mediated diseases (Nature Communications, 2024)
• PCSK9 loss-of-function variants and coronary heart disease (NEJM, 2006)
• FOURIER trial of evolocumab (NEJM)

Biology databases and tools

• Reactome
• KEGG
• Gene Ontology
• STRING
• Open Targets
• ChEMBL
• DrugBank

Pangenome resources

• Human Pangenome Reference Consortium
• NHGRI Human Genome Reference Program
• HPRC Data Release 2
• HPRC release timeline
• The 0 Genome

Other links cited in the text

• dimenwarper on X
• Bryan Johnson on X
• Anne Wojcicki on X
• oak.bio