Playing the game vs. finding a cheat code

Pokémon and cell types

Imagine a new Pokémon game has just come out, and you really want to catch a Zapdos. It’s listed in the game’s Pokédex, so you know it must be possible to catch, but you’re not sure how.

You could either:

1. Play the game normally. There are some hints that Zapdos like to hang out in thunderclouds so you could try looking there.

2. Randomly mash buttons and look for any weird glitches which you can exploit to give yourself a Zapdos.

Zapdos is a legendary lightning-bird Pokémon that looks pretty cool.

Advantages of playing normally

• You know it’s possible to get a Zapdos this way, it will just take a while and you’ll need to explore thunderclouds a lot.

• By exploring thunderclouds, you might learn more about them.

• If you get a Zapdos using a glitch, it’s possible that the save file might be corrupted in some way and your Zapdos won’t work quite right.

Advantages of button mashing

• Other players have found glitches to get other Pokémon, and although nobody has got a Zapdos yet, it seems plausible that it could work.

• By finding glitches, you might learn more about the internal logic of the game.

• If you have the code of the game to examine, you can more easily find glitches.

• Once you’ve found the proper sequence of button mashes, it’s a lot easier to get a Zapdos than by playing the game. This is good if you want to get more of them, or if you want to share the cheat code you’ve found with other people.

Gotta grow ‘em all

Now, I don’t care so much about catching rare Pokémon, but I do really want to grow an oocyte in cell culture. Again, I have two choices for my general approach:

1. Try to mimic the developmental signaling processes of the ovary, going from pluripotent stem cells, to primordial germ cells, to oogonia, to oocytes.

2. Use transcription factors to reprogram the cells into oocytes (including activating meiosis).

Both approaches are valid!

The first approach has been very successful in mice, shown in several groundbreaking papers¹ by the labs of Mitinori Saitou and Katsuhiko Hayashi. Although it will certainly be more challenging in humans due to the longer time required for development, I think it’s promising. For this to work, a strong knowledge of human ovarian development is key, and until recently, this was not available due to lack of data on human fetal ovaries.

The second approach has also been somewhat successful. The Hayashi lab has made mouse oocyte-like cells from embryonic stem cells using transcription factor (TF) expression², although these cells did not properly do meiosis. More recently, I used TF expression to generate functional human granulosa-like cells and oogonia-like cells from stem cells.³ I also have some encouraging preliminary data about activating meiosis-related genes using transcription factors. And of course, let’s not forget that many other cell types⁴ (including pluripotent stem cells) can be generated by TF expression. For this approach, powerful screening methods are required to identify candidate TFs and quickly test many TF combinations.

The advantages and disadvantages for each approach are similar to what I explained above for the Pokémon metaphor. We know it’s possible to grow oocytes following normal developmental pathways, since nearly all women do so. But, following the normal pathways would take at minimum 9 months (and up to ~14 years if strictly following them). This is much slower than in mice, where only a few weeks are necessary. Using TFs can definitely speed things up, but the epigenetic quality of the resulting cells will be challenging to get right. Of course, a combination of both approaches is also possible: use natural signaling pathways to activate some stages of ovarian development, and TF expression (or other synthetic tools) to skip over other stages.

More broadly . . .

I think the “playing the game” vs. “finding/using a cheat code” distinction is useful in a variety of other areas:

• Natural product manufacturing: enzymes vs. organic synthesis⁵

• Evaluating computer code: static analysis vs. fuzzing

• Weight loss: eating less vs. taking medication

• Treating sickle cell disease: correcting the mutation vs. making cells switch to fetal hemoglobin.

• Imaging: superresolution microscopy vs. expansion microscopy

• Paying your debt vs. trying to get it canceled through the courts

• Emulating a human brain vs. building an AI with its own architecture

Note that combined approaches (for example, semisynthesis) are also possible for most of these examples.

If you have any other examples, post them below!

1. Papers: 1, 2, 3

2. Hamazaki et al. 2020

3. Papers: 1, 2

4. For example: the TFome project and the MORFeome project.

5. For example: compare vitamin B12 vs. vanillin.