My impression from reading The Computer Scientist’s Guide to Cell Biology is that the reason for woefully inadequate descriptions of cellular systems, rather than any fundamental deficiency on the part of biologists, is that the systems are so dang tiny and crowded, with every molecule running into every other molecule every second or so, at sizes literally below the widths of light waves, that we just don’t have the tools to physically observe what’s happening in real-time. Even the primitive equivalents of “shooting individual components” took so much clever innovation and hard work to figure out how to do that frankly it’s amazing that we know everything that we do.
There are interesting attempts at modeling cells using Boolean networks (for example inferred using SAT/SMT), differential equations (preferably stochastic) and also with process algebra. Some are pretty successful.
> [...] we don’t have the tools to physically observe what’s happening in real-time
We do have single-cell technology, which is pretty close as you can get thousands of snapshots of different cells from the same lineage. One can then reorder them into some sort of pseudotime. There are also interesting real-time reporting systems for bacterial cells, but they are not high throughput.
I think the field is still very young, and the current generation of biology professors is mostly allergic to formalism. Some departments are not, a notable exception are Caltech and Cold Spring Harbor (where the author of the essay is based).
Single cell sequencing technology is analogous to taking many radios, and listing all the parts the radios contains. Then clustering radios by similarity in their part abundances.
Useful, but a parts list is still far from figuring out how a radio works.
>is that the systems are so dang tiny and crowded, with every molecule running into every other molecule every second or so,
Per my two earlier comments [1], I would phrase it as an issue of biological systems optimizing soley for fitness, while (human-designed) computers are also heavily optimized for intelligibility, modularity, and ease of reasoning about. This makes it much easier to isolate and experiment with the subsystems and gain an understand, in contrast to biological systems, which will constantly bleed state across the entire system.
> Another argument is that we know too little to analyze cells in the way engineers analyze their systems. But, the question is whether we would be able to understand what we need to learn if we do not use a formal description. The biochemists would measure rates and concentrations to understand how biochemical processes work. A discrepancy between the measured and calculated values would indicate a missing link and lead to the discovery of a new enzyme, and a better understanding of the subject of investigation.
> Do we know what to measure to understand a signal transduction pathway? Are we even convinced that we need to measure something? As Sydney Brenner noted, it seems that biochemistry disappeared in the same year as communism (Brenner, 1995
). I think that a formal description would make the need to measure a system's parameters obvious and would help to understand what these parameters are.
Maybe it’s that I don’t grok what the author means by “formal description” but I don’t see how that really touches on it.
As a layman the impression I got from the article is “biologists are doing the wrong things to learn about cells, I think they should be doing this other thing instead” whereas the impression I have from what I’ve read about those learning processes is that the learning is so hard that biologists are literally just doing whatever they can come up with to learn anything at all. They didn’t choose these processes because of what they wanted to learn, they literally invented those processes because it was what they could come up with (pretty ingeniously, too), there are no alternatives until those are invented and they would be ecstatic to use them as well.
I probably have a disconnect from my lack of knowledge. An example would help; if it’s in the radio analogy I’ve lost it by the orders of magnitude difference in merely observing it at all.
The radio schematic in the paper looks kind of unusual, with a dual-gate MOSFET. Also a bunch of varactor diodes for tuning. Any radio experts want to comment on this circuit? I think it's FM, but I'm not sure.
I don't think that's a radio, just an antenna amplifier. I'm not seeing where demodulation would happen. Some values seem kinda weird to me, like R4 on top of the cascode seems sorta high for something that oughta have a lot of bandwidth (and the way L2+C6 are drawn is odd). Likewise 390k on the emitter of Tr3, seems like very little room for bias for something that's supposed to drive 50 ohms. Bandwidth at the input looks to be <2 MHz, so maybe it's meant for AM broadcast or 160m.
R4 controls the output impedance of TR1. It'd have to be high to keep the loaded Q of the L2 tank high. R10 looks like a mistake to me. Probably should be 390 ohms.
Agree with your assertions though.
It might have been reverse engineered from something badly or drawn up by someone smoking crack. Even Tektronix had that problem on occasions in their service manuals ...
330uH is a pretty big inductor for FM/VHF. RV1 controls RF gain (think of TR1 as two JFETs, one a current source and one a common source amplifier in a small signal model). TUNE VOLTS changes the resonant frequency of the two tank circuits. TR2/TR3 is an impedance buffer. The rest is decoupling.
Looks like a tunable preselector to me rather than a receiver. So part of a receiver yes.
Previous discussions:
June 2022: https://news.ycombinator.com/item?id=31697757 (22 comments)
January 2022: https://news.ycombinator.com/item?id=30120457 (19 comments)