Computational Biology

[J_B_]

This may seem a strange question to ask, but is anyone here familiar with computational biology, specifically building population genetics models of evolutionary systems from first principles?

 

[The Barbarian]

Years ago, I was getting a degree in systems at USC, and one of the projects was to apply the Lotka-Volterra model to insular predator/prey biomes.

Pretty primitive stuff compared to what's being done today. I'm interested in what you know about that kind of modeling applied to genetics.

 

[J_B_]

Years ago, I was getting a degree in systems at USC, and one of the projects was to apply the Lotka-Volterra model to insular predator/prey biomes.

Pretty primitive stuff compared to what's being done today. I'm interested in what you know about that kind of modeling applied to genetics.

Almost nothing.

Though I spent my career in mechanical engineering, I've had a talent for computer science ever since I was a kid. As such, it's a hobby of mine to dabble in how all kinds of different areas use computers.

For a long time biology held no interest for me, but one of my kids ended up in the medical field. That prompted me to start playing with computational biology. I've started looking at some code. That's the easy part. Grasping the in's and out's of the science is the more challenging part, though I'm having fun with it.

The question came simply because I've been digging through some material recently.

 

[jhwatts]

I took Mathematical Biology as a graduate course and Biophysics. Why?

 

[J_B_]

I took Mathematical Biology as a graduate course and Biophysics. Why?

I was interested in learning more about it.

For example, in population dynamics is the code structure centered primarily around empirical statistics, or are models built up from first principles?

 

[sfs]

I'm a computational biologist and I've done a fair bit of population genetics modeling. I'd say models cover a wide range, depending on the application. All models require some empirical input and all have some theoretical framework that provides the structure. Simple heuristic models can be pretty close to being first principles-based. Bigger, more sophisticated models tend to incorporate a lot of both.

 

[J_B_]

I'm a computational biologist and I've done a fair bit of population genetics modeling. I'd say models cover a wide range, depending on the application. All models require some empirical input and all have some theoretical framework that provides the structure. Simple heuristic models can be pretty close to being first principles-based. Bigger, more sophisticated models tend to incorporate a lot of both.

Thanks. That's a good summary. I could deluge you with a lot of questions stemming from this, but it seems rather impolite to impose too much on your time. Probably my top 3 questions would be:
* Are there a few (or one) software tools that are most commonly used?
* What first principles do they employ?
* What are the major issues that lead biologists to move away from a purely first principles approach toward empirical data? ... I assume there are people working to improve model accuracy.

 

[The Barbarian]

I'm a computational biologist and I've done a fair bit of population genetics modeling.

Is there decent introductory text for a somewhat outdated biologist to catch up on?

 

[sfs]

Thanks. That's a good summary. I could deluge you with a lot of questions stemming from this, but it seems rather impolite to impose too much on your time. Probably my top 3 questions would be:
* Are there a few (or one) software tools that are most commonly used?

Not really. There is such a wide range of applications that require different tools, plus lots of researchers want to invent their own tool. Somebody studying the evolution of SARS-CoV-2 under the influence of immunity is going to be looking at different kinds of models than somebody designing rare variant disease association studies, or looking for migration and admixture among human ancestors, or inferring the variation in ancestral population size in some species, or identifying parts of the genome under recent positive selection.

What first principles do they employ?

Standard biological processes: reproduction, mutation, recombination (where appropriate), selection, etc.

What are the major issues that lead biologists to move away from a purely first principles approach toward empirical data?

Biology in general doesn't lend itself well to trying to extrapolate first principles into the behavior of specific systems: the systems are just too complex. Any realistic model you can generate from first principles is going to have an enormous number of free parameters that have to be estimated from empirical data (or sometimes just guessed) and even the structure of the model likely has to be chosen to fit the specifics of the population you're interested in.

 

[sfs]

Is there decent introductory text for a somewhat outdated biologist to catch up on?

I don't know of any but I'm not a great person to ask, since I don't teach and I learned the field on the job, so to speak. There are introductory textbooks on population genetics as a whole (Hartl and Clark is one standard one -- and I know one of the authors), but there may be too many specific applications for pop gen models for easy summary.

 

[J_B_]

Not really. There is such a wide range of applications that require different tools, plus lots of researchers want to invent their own tool. Somebody studying the evolution of SARS-CoV-2 under the influence of immunity is going to be looking at different kinds of models than somebody designing rare variant disease association studies, or looking for migration and admixture among human ancestors, or inferring the variation in ancestral population size in some species, or identifying parts of the genome under recent positive selection.

I suspected this was the case, but it's good to get confirmation. I did a quick survey of the literature and noticed a lot of people promoting their tool. It looked like there were a few attempts to create an open source repository, but I don't see that it's caught on. Too bad. It's a good way to pool resources and build applications from common code - a hive mind of sorts.

Biology in general doesn't lend itself well to trying to extrapolate first principles into the behavior of specific systems: the systems are just too complex. Any realistic model you can generate from first principles is going to have an enormous number of free parameters that have to be estimated from empirical data (or sometimes just guessed) and even the structure of the model likely has to be chosen to fit the specifics of the population you're interested in.

I can believe that's the case. My engineering models sometimes seem near the limits of what the industry can support in terms of complexity, and I imagine living systems are magnitudes more complex.

Standard biological processes: reproduction, mutation, recombination (where appropriate), selection, etc.

Even though I accept what you said above, it would still be fun to pick out one of these and try. I once tossed out a few ideas to my kid who's in the medical field. All I got was a laugh and a comment that I was approaching biology like an engineer.

 

[BobRyan]

I'm a computational biologist and I've done a fair bit of population genetics modeling. I'd say models cover a wide range, depending on the application. All models require some empirical input and all have some theoretical framework that provides the structure. Simple heuristic models can be pretty close to being first principles-based. Bigger, more sophisticated models tend to incorporate a lot of both.

Do you do anything with mutation rates and limits or bounds on it?

Sandwalk: Calculating time of divergence using genome sequences and mutation rates (humans vs other apes).

It appears human mutation rates are very high

Mutation rate - Wikipedia

"The human mutation rate is higher in the male germ line (sperm) than the female (egg cells), but estimates of the exact rate have varied by an order of magnitude or more. This means that a human genome accumulates around 64 new mutations per generation because each full generation involves a number of cell divisions to generate gametes.[12] Human mitochondrial DNA has been estimated to have mutation rates of ~3× or ~2.7×10−5 per base per 20 year generation (depending on the method of estimation);[13] these rates are considered to be significantly higher than rates of human genomic mutation at ~2.5×10−8 per base per generation.[14] Using data available from whole genome sequencing, the human genome mutation rate is similarly estimated to be ~1.1×10−8 per site per generation.[15]"

 

[The Barbarian]

Rob, do you know what mitochondrial DNA is?

Mitochondria are endosymbiotic bacteria, which is why they have bacterial DNA, which is different than eukaryotic DNA. And bacterial DNA is much more likely to be variable in mutation rates than eukaryotic DNA:

Evolution of mutation rates in bacteria
Evolution of mutation rates in bacteria - PubMed

For our own genomes, however:

Every time human DNA is passed from one generation to the next it accumulates 100–200 new mutations, according to a DNA-sequencing analysis of the Y chromosome.

This number — the first direct measurement of the human mutation rate — is equivalent to one mutation in every 30 million base pairs, and matches previous estimates from species comparisons and rare disease screens.

The British-Chinese research team that came up with the estimate sequenced ten million base pairs on the Y chromosome from two men living in rural China who were distant relatives. These men had inherited the same ancestral male-only chromosome from a common relative who was born more than 200 years ago. Over the subsequent 13 generations, this Y chromosome was passed faithfully from father to son, albeit with rare DNA copying mistakes.

The researchers cultured cells taken from the two men, and using next-generation sequencing technologies found 23 candidate mutations. Then they validated twelve of these mutations using traditional sequencing techniques. Eight of these mutations, however, had arisen in their cell-culturing process, which left just four genuine, heritable mutations. Extrapolating that result to the whole genome gives a mutation rate of around one in 30 million base pairs.
Human mutation rate revealed - Nature

So each of us has about 100 mutations that weren't present in either of our parents. Not all of these end up spreading through human populations, of course, but many of them do. The vast majority of them don't do much of anything, but some of those might be very slightly harmful, even harmful enough for natural selection to remove them.

 

[sfs]

Even though I accept what you said above, it would still be fun to pick out one of these and try. I once tossed out a few ideas to my kid who's in the medical field. All I got was a laugh and a comment that I was approaching biology like an engineer.

It's not exactly evolution, but colleagues at the Institute for Disease Modeling have created very complex agent-based models of disease transmission that include lots of details about multiple processes. Tuning them is a challenge, as is generalizing results generated under the conditions present in one location.

 

[sfs]

Do you do anything with mutation rates and limits or bounds on it?

When I worked in human genetics (which I left behind for infectious disease research a few years ago), I used estimates of the mutation rate but didn't contribute to making them.

 

[BobRyan]

Rob, do you know what mitochondrial DNA is?

Mitochondria are endosymbiotic bacteria, which is why they have bacterial DNA, which is different than eukaryotic DNA.

I was not aware that an organelle like mitochondria is classified as bacteria. Rather I think there are some guesses that Mitochondria used to be bacteria. As far as I know they are not bacteria if we just consider observations in nature. What we actually see under a microscope.

 

[The Barbarian]

I was not aware that an organelle like mitochondria is classified as bacteria. Rather I think there are some guesses that Mitochondria used to be bacteria.

Evidence, not guesses. You see, mitochondria have circular DNA like other bacteria. And their genome is most like that of bacteria:

Molecular biologists have discovered how ancient bacteria gradually evolved into mitochondria in eukaryotic cells1. Those eukaryotic cells ultimately gave rise to all complex life forms, such as plants and animals.

The researchers identified two key changes in the protein synthesis and surveillance machineries of the eukaryotic cells that resulted in the successful transition from bacteria to mitochondria.

Amino acids come in two different forms – L and D. L-amino acids are predominantly used to make proteins in cells. From bacteria to mammals, cells use D-aminoacyl-tRNA deacylase (DTD), a proofreading enzyme that only allows L-amino acids, to prevent D-amino acids being incorporated into proteins.

The research team, led by Rajan Sankaranarayanan at the Centre for Cellular and Molecular Biology (CCMB) in Hyderabad, found that the bacterial DTD had undergone a change that was compatible for use in eukaryotic cells. They also found a change in the discriminator element for mitochondrial transfer RNA (tRNA) that shuttles glycine, an essential amino acid for making a protein.

Of the 1,000 proteins identified in mitochondria, 40% are of bacterial origin. To trace back mitochondria’s origin, the scientists did experiments with Jakobida, the earliest known, tiny free-living organisms that retain a bacteria-like mitochondrial genome.
How mitochondria evolved from bacteria

But is there any proof that such an endosymbiotic relationship can form? Turns out, there is:
Trends Cell Biol 1995 March5(3) 137-40
Bacterial endosymbiosis in amoebae
K W Jeon
Abstract

The large, free-living amoebae are inherently phagocytic. They capture, ingest and digest microbes within their phagolysosomes, including those that survive in other cells. One exception is an unidentified strain of Gram-negative, rod-shaped bacteria that spontaneously infected the D strain of Amoeba proteus and came to survive inside them. These bacteria established a stable symbiotic relationship with amoebae that has resulted in phenotypic modulation of the host and mutual dependence for survival.

 

[J_B_]

It's not exactly evolution, but colleagues at the Institute for Disease Modeling have created very complex agent-based models of disease transmission that include lots of details about multiple processes. Tuning them is a challenge, as is generalizing results generated under the conditions present in one location.

Thanks for letting me know. I did a quick check, and it looks like they have shared an open source code called EMOD.