Biology needs better visualizations

Who will be our Bartosz Ciechanowski?

Gastrulation is one of the most fundamental processes in developmental biology, but up until a few years ago, I didn’t really understand it.¹ How does cell differentiation relate to the morphological changes in the embryo? And why are the embryos of some species (like mice) shaped like cylinders during this stage, whereas others (like humans) are shaped like disks?² The main barrier to my understanding was the lack of a mental model of how the different embryonic cell types are organized over time. Journal articles and textbooks had diagrams, and there were a few Youtube videos of chicken embryos, but none of these really showed the whole process in full detail.

What I really wanted was an interactive 3D movie of a developing embryo – one where I could pause it at a particular stage, zoom in and look at a group of cells, and then see how they moved and grew when I un-paused it. The things that come closest to this are the eMouse Atlas, the Virtual Human Embryo, and the 3D Embryo Atlas, but these websites just show 3D reconstructions of different embryonic stages (not integrated videos), and they aren’t great about labeling the different cell types.

If you haven’t yet seen Bartosz Ciechanowski’s visualizations of a car engine or a mechanical watch, set aside this post, take a look at his work, and come back when you’re done. Trust me, they’re super cool.

This is just a screenshot, you’ll need to visit the page for the full experience.

What I really want is a Ciechanowski-like visualization of the developing embryo. Ideally this could also show gene expression patterns and cell signaling,³ and things like morphogen gradients. So why don’t we have this?

Lack of data?

Historically, this was a problem. Mammalian development happens inside the mother’s uterus, so it’s pretty hard to see it happen in real time. Still, analysis of preserved embryonic tissues (such as the Carnegie collection of human embryos) provided detailed morphological information, allowing the identification and classification of developmental stages. For other organisms like worms, flies, and frogs, tracking embryonic development was even easier.

And even in the 1990s, methods like RNA in situ hybridization could map gene expression patterns in the embryo. Today, biologists can measure the whole transcriptome at single-cell or even sub-cellular resolution. All those fancy single-cell and spatial transcriptomics methods must be good for something, right?

In situ hybridization of hunchback mRNA in a fly embryo (source)

Although no single data source contains all the information necessary to make an interactive embryo visualization, I believe that lack of data is not an issue here.

Animators and biologists rarely collaborate

The real issue is that making an interactive embryo visualization would require deep expertise in both biology (to integrate all the data) and 3D graphics, animation, and UI design (to display it). Traditional forms of academic publishing, such as journals and textbooks, focus on 2D images, and even when they include movies, they’re inherently non-interactive. Therefore, biologists are rarely familiar with other forms of visualizations. And graphic designers and animators aren’t typically included in biology labs, because it’s hard to justify spending limited grant money on paying their salaries when they aren’t directly generating new biological insights.

The end result is that while huge sums of money are being spent on collecting data, the full potential of the data isn’t realized for lack of good visualizations.

How can we change this?

Basically, we need to get someone like Bartosz Ciechanowski⁴ to team up with a developmental biology research group. On a somewhat broader scale, this would be a good topic for a Focused Research Organization, which could fund the production of visualizations of various biological systems – not just embryonic development, but also things like the brain, the chromosomes during meiosis, or the ribosome translating a protein. Each visualization would be a comprehensive, interactive map of a biological process as it progresses over time, with the parts labeled.

Better visualizations will make it easier to train new researchers, and will also increase the productivity of skilled researchers by helping them to reason about biological processes. Thinking about a mechanical watch is so much easier if you can actually see the gears. Overall, increasing the funding available for high-quality visualizations will substantially accelerate the pace of scientific progress.

1. I’m not sure I can give a simple explanation of gastrulation, but here’s my best attempt: In the early embryo, the epiblast is made of stem cells that have the potential to form any cell type in the adult body. Gastrulation is when the epiblast stem cells commit to forming particular cell lineages. This is accompanied by morphological changes in the embryo, but these changes vary by species. The first image in the Wikipedia article is a bit misleading: that’s how gastrulation looks in sea urchins and amphibians, but bird and mammal embryos are shaped differently.

And it’s not just me who struggled: pre-meds find gastrulation confusing too. (1, 2). Whole textbooks have been written on the process.

2. This paper, published in 2021, provides an answer: the difference is not anything in the embryo itself, but instead it’s caused by different physical forces exerted by the trophectoderm. Which raises the question: why are those forces different? As far as I can tell this is still unknown.

3. For example, I could select a gene from a list, and the cells would be colored according to its expression.

4. I actually asked him about doing an embryo visualization in 2022, but he said he gets too many requests to make visualizations, so he only does the ones he wants, not ones which other people suggest.