Same Inflammation, Two Shutter Speeds

Anyone who has photographed the night sky knows a fundamental truth: shutter speed determines what you can see.

At 1/1000 second, you capture a few bright stars amid overwhelming noise. At 30 seconds, star trails emerge and the Milky Way reveals itself. Same sky, same lens — the only variable is exposure time. Yet the difference in information between the two images is staggering.

I recently read a paper that made me realize: the way we observe concepts in biomedicine is exactly like photographing the night sky. Get the shutter speed wrong, and the signal drowns in noise.

A Stunning Number: 6.4x

In 2021, Conole and colleagues published a study in Neurology. They examined how two different representations of the same concept — chronic inflammation — correlated with brain health outcomes.

The first: serum CRP protein concentration. This is the most widely used clinical inflammation marker. Draw blood, measure the protein, get a result. Every hospital in the world does this.

The second: DNAm CRP — a DNA methylation surrogate for CRP. Rather than measuring the protein directly, it reads the inflammatory "signature" left on the genome through a set of methylation sites.

They then calculated how each representation correlated with brain volume:

| | Serum CRP | DNAm CRP |

|---|---|---|

| Correlation with brain volume | β ≈ -0.03 | β ≈ -0.20 |

| Focus state | Out of focus | In focus |

Same concept. Different representation. Correlation improved by 6.4x.

6.4 times. Not 10%, not 20% — 6.4 times.

Why Such a Dramatic Difference?

The answer lies in a question I have been pondering for years: what are we actually measuring?

Serum CRP protein concentration measures inflammation right now. Draw blood in the morning versus afternoon, and the result can differ by several-fold. Didn't sleep well last night? CRP goes up. Had a cold last week? CRP hasn't come back down. Just finished a run? CRP changed again. It is an extremely sensitive "snapshot" — a shutter speed of 1/1000 second.

The problem: brain atrophy doesn't happen in a minute. It is the cumulative result of chronic inflammation over years, even decades. Using a 1/1000-second shutter to capture a decade-long process will inevitably produce a blurry image.

DNAm CRP is fundamentally different. DNA methylation changes slowly — on timescales of weeks, months, years. When multiple methylation sites are trained to predict CRP levels, they are effectively performing a temporal integration of the inflammatory signal. They are not asking "what is your inflammation level today?" but rather "how inflamed have you been over the past few years?"

This is long exposure.

The shutter speed went from 1/1000 second to 30 seconds. The star trails appeared.

It's Not That Methylation Is "Better" — Its Timescale Simply Matches

I want to emphasize one point: it's not that methylation is inherently more "advanced" than protein. It's that for the specific property of "chronicity," methylation's temporal resolution happens to match.

If we were studying an acute concept — say, the immune response 2 hours after drug injection — then protein concentration might actually be the better "focal length." Because the rate of protein change matches the timescale of acute processes.

The key is not which omics layer is "better." The key is: the dynamic timescale of the concept must match the temporal resolution of the representation medium. Match, and you're in focus. Mismatch, and you're blurred.

It's like needing different shutter speeds for different subjects:

- Flying birds: 1/2000s (metabolites, second-level changes)

- Walking people: 1/125s (proteins, hour-level changes)

- Flowing water: 1/4s (transcriptome, day-level changes)

- Star trails: 30s (DNA methylation, month-year-level changes)

- The genome itself: no shutter needed (DNA sequence, unchanging)

Chronic inflammation is a star trail. Using a bird-photography shutter to capture star trails yields nothing but darkness.

Prisms, Focal Lengths, and EvoSika

In my previous essay, I described how EvoSika treats every biomedical concept as an object that needs to be "focused." The concept itself is real, but its clarity varies dramatically across different representation media.

Conole's paper is the perfect empirical validation of this framework.

Chronic inflammation is the concept. CRP is the prism — a lens through which the concept is refracted. Serum protein concentration is one focal length; methylation signature is another.

Same prism, same concept. Change the focal length, and clarity differs by 6.4x.

EvoSika's mission is to systematically find the optimal focal length for every concept. Not relying on intuition to choose an omics layer or a gene set, but using a computational framework to automatically match a concept's timescale with the temporal resolution of its representation medium.

A Deeper Implication: Omics Layers Are Not Parallel — They Are Sequential

This finding also suggests something more profound.

Traditionally, we think of the proteome, transcriptome, and methylome as three "parallel perspectives" — like photographing the same building from different angles. But Conole's paper reveals: they are not parallel perspectives, but integrations over different timescales.

Proteins are short exposures. Methylation is long exposure. They are not "different angles" — they are "different shutter speeds."

This means:

For "chronic" concepts (chronic inflammation, aging, neurodegeneration), slower omics layers focus better.

For "acute" concepts (acute drug responses, stress reactions), faster omics layers focus better.

Different concepts require different "exposure times" to come into focus. This is not a technical detail. This is a fundamental principle of biological knowledge representation.

Epilogue

After twenty years in computational biology, I am increasingly convinced of one thing: the biggest problem in biology is not a lack of data — it's that we 'look' at data the wrong way.

The 6.4x improvement in Conole's paper is not an isolated case. It reveals a universal principle: when you choose the correct representation medium for a concept — the correct shutter speed — the signal emerges from the noise.

This is not a technical optimization problem. This is an epistemological one.

What EvoSika is building is an auto-focus system for every concept in biomedicine. Not asking scientists to guess which omics layer or gene set might work better, but letting the computational framework find the optimal focal length on its own.

When the shutter speed is right, you see — for the first time — what you've been trying to see all along.

Xiong Jianghui

Founder, EvoSika