Chapter 19: Inheritance of Complex Traits

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replaced the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

Welcome to the Deep Dive.

Today, we're tackling something a bit more complicated than your standard genetics lesson.

We're looking at the traits that don't follow those neat Mendelian rules.

Exactly.

Things like height, weight, maybe even personality or your risk for certain diseases.

These aren't simple on -off switches.

Right.

Think about this scenario.

Paul Reston, he's 47, a teacher, starts getting chest pains.

His dad died young, heart attack of 45.

His Paul just, you know, faded to follow.

And his doctor would say, well, it's not that simple.

It's your diet, your exercise habits, smoking history, and probably a whole bunch of genes interacting.

It's that classic nature plus nurture situation.

Precisely.

And that's our mission for this Deep Dive.

We're digging into these complex traits.

We need the tools, statistical and molecular, to understand them.

So what are we hoping you take away from this?

Well, first, what are quantitative traits?

How did scientists figure out you could separate genetic and environmental effects?

Okay.

And then the math part.

Yeah.

How we use statistics like variance to calculate something called heritability.

It's crucial.

And finally, the modern stuff.

Like how we actually find the specific genes involved using things like QTL mapping and GWAS.

Exactly.

From basic concepts to cutting edge techniques.

All right.

Let's get our term straight.

A complex trait that's influenced by multiple genes, right?

Polygenic.

Yeah.

And the environment.

Correct.

But to study them scientifically, we often need something measurable, something we can put a number on.

And those are quantitative traits.

Yep.

Think height, weight, blood pressure.

Things that vary along a continuum.

You can see the genetic influence just by looking at families, can't you?

Especially identical twins.

They tend to be much more similar for these traits than fraternal twins.

Makes sense.

And we also see it in agriculture with selective breeding.

Farmers wouldn't be able to breed for, say, higher milk yield if there wasn't underlying genetic variation to select from.

Absolutely.

That's strong practical evidence.

But the scientific groundwork really started getting laid in the early 20th century.

Wilhelm Johansson in Denmark, working with beans.

Beans?

What did beans tell us?

He studied broad bean seed weight.

He bred them until he had what he called pure lines,

basically.

Genetically identical plants.

Okay.

So they all have the same genes.

Right.

But interestingly, their seeds still varied in weight.

Since the genes were the same within a pure line, that variation had to be caused by subtle environmental differences.

Ah, so he experimentally separated genetic uniformity from environmental effects.

Exactly.

Proved that the phenotype, what you actually see and measure, is always a mix.

P equals G plus E.

Phenotype equals genotype plus environment.

Foundational stuff.

So Johansson showed G and E are both players.

Then how do we figure out the multiple genes part, the polygenic inheritance?

That takes us to Herman Nilsson L in Sweden looking at wheat grain color.

This is a classic example.

Wheat color.

Okay.

He crossed a red -grained wheat with a white -grained one.

The first generation, the F1, they were all intermediate red.

Kind of makes sense, like mixing paint.

But the F2 generation.

That's where it got cool.

He didn't just get red and white back.

He got a whole spectrum seven distinct color classes.

From white all the way to the original dark red.

Seven classes.

How?

He figured the only way to explain that pattern was if three separate genes controlled the color and each gene had alleles that either added a dose of red pigment or added none.

So like zero doses gives white and six doses two for each of the three genes gives dark red.

Precisely.

And everything in between.

It showed multiple genes acting together additively.

And Edward East saw something similar with tobacco flower length.

Yeah.

Similar idea, but with a truly continuous trait.

He crossed long flowered and short flowered tobacco.

Again, F1 intermediate.

But his F2 generation showed way more variation than either parent or the F1.

So the genes were segregating, shuffling around, creating all these new combinations.

Plus you still had that environmental wobble Johansson showed.

Exactly.

East also tried to estimate how many genes were involved.

He looked at how rare it was to get back to the original parental extremes in the F2.

He didn't find any in over 400 plants.

Which meant?

Which meant it was statistically unlikely to be just a few genes.

He calculated it had to be at least five genes contributing to flower length.

A crude method, but it pointed towards polygenic control.

Okay.

Height, weight, flower length.

They vary smoothly.

But what about things like, say, cleft lip or back to Paul's heart disease risk?

You either have it or you don't.

How is that quantitative?

Good question.

Those are called threshold traits.

They look discontinuous, yes or no.

But the idea is there's an underlying continuous variable we can't see called liability.

Liability, like your overall risk.

Sort of, yeah.

It's influenced by multiple genetic factors, environmental factors.

Think cholesterol, blood pressure, genetic markers adding up.

And only when your total liability crosses a certain threshold does the condition actually appear.

Like a tipping point.

Exactly.

Heart disease, schizophrenia, cleft lip.

They often work like this.

And twin studies back this up.

Definitely.

We look at concordance rates.

What percentage of twin pairs both have the trait.

For identical monozygotic MZ twins, the concordance for something like cleft lip is around 40%.

And for fraternal dizygotic DZ twins.

Only about 4%.

That huge difference points to a strong genetic component influencing that underlying liability.

Schizophrenia shows a similar pattern.

MZ concordance much higher than DZ.

Okay, so these traits are complex, often continuous, or based on an underlying continuous liability.

We can't just count P's.

We need stats.

We absolutely do.

The first step is just describing the data.

You get a bunch of measurements, heights, weights, whatever.

You need to summarize them.

So you calculate the average, the mean, that tells you the center of the distribution.

Right.

But just knowing the average isn't enough, are all the measurements clustered tightly around the mean or are they really spread out?

That spread is measured by the variance.

Variance involves squaring differences from the mean, right?

Makes bigger differences count more.

Exactly.

It quantifies the dispersion.

And if you take the square root of the variance, you get the standard deviation.

That's useful because it's back in the original units, like inches, pounds, or whatever you measured.

And for a normal bell curve, most data falls within a couple of standard deviations of the mean.

About 95 % within two standard deviations, yes.

So mean and variance, or standard deviation, give you a good snapshot of the data.

But the real breakthrough was figuring out how to connect this to genetics and environment.

That was R .A.

Fisher, wasn't it?

Around 1918.

Yes, a giant in the field.

Fisher formalized the multiple factor hypothesis.

He basically said, look, any individual's phenotype is the overall population mean plus some deviation due to their specific genes and some deviation due to their specific environment.

Mu plus G plus EU.

Simple equation, but profound implications.

Hugely profound.

Because he then showed mathematically that the variance of the phenotypes, VDTV, could be broken down.

It could be partitioned into the variance caused by genetic differences and the variance caused by environmental differences.

So VTE if you list plus V.

Ah, separating the sources of variation.

Yeah.

Like Johansen did experimentally, but now with statistics for a whole population.

Precisely.

This equation is the bedrock of quantitative genetics.

Can we walk through that wheat maturation time example?

How does the partitioning actually work with numbers?

Sure.

So in that study, they had genetically uniform lines, the two parent strains A and B, and their F1 hybrid.

Since they're genetically identical, any variation in their maturation time is assumed to be purely environmental variance, fee dollars.

Okay.

So you measure the variance within each of those uniform groups and average them.

Yeah.

And let's say the average variance, our estimate of dollars comes out to 2 .28 days squared.

Okay.

That's the environmental noise level.

Right.

Then you look at the F2 generation.

These plants are genetically diverse because of their genetic segregation.

Their total phenotypic variance V.

Tyler reflects both genetic differences and environmental effects.

Let's say V dollars for the F2 is measured at 14 .286 days squared.

So now we have V dollars in values.

We just subtract V dollars equals $14 and $26, which equals 1198 days squared.

That number 1198 is our estimate of the variance specifically due to genetic differences among the F2 plants.

Okay.

So we've split the total variance into a genetic part and an environmental part.

V dollar.

Now, how much does the genetic part actually matter?

That sounds like heritability.

Exactly.

The first measure we calculate is broad sense heritability written as 2H202.

It's simply the proportion of the total phenotypic variance that's due to genetic variance.

2H2 week is VDVTE.

So for the wheat example, that would be 11 .98 divided by 14 .26.

Right.

Which comes out to about 0 .84.

So we'd say 84 % of the variation in maturation time in F2 population is attributable to genetic differences.

84 % sounds pretty high.

Does that mean we can easily predict maturation time of offspring?

Not necessarily.

That's the catch with broad sense heritability.

It includes all types of genetic effects.

But for prediction, especially in breeding, we need something more specific.

Why?

What other kinds of genetic effects are there?

Well, V dollars itself can be subdivided.

There's additive genetic variance, which is the straightforward effect of alleles adding up like in the wheat color example.

But there's also variance due to dominance, where one allele masks another, and epistasis, where genes at different loci interact.

And dominance and epistasis mess up predictions.

They do because they make the relationship between genotype and phenotype nonlinear less predictable.

Dominance hides recessive alleles, for instance.

Additive effects, however, are reliably passed from parent to offspring.

So we need a measure that only considers the additive part.

Exactly.

That's Narasen's heritability, written T1 dollars.

It's defined as T1 dollars equals vey Vt.

It's the proportion of total variance due solely to additive genetic effects.

This is the heritability that breeders really care about.

Let's make that concrete.

Additive versus dominance.

Okay.

Think Snapdragon flower color.

Red allele adds pigment, white adds none.

The heterozygote is perfectly pink intermediate.

That's additive.

Each allele contributes predictably.

Now think ABL blood types.

An AO genotype gives type A blood, just like AA.

The dominance of A over O obscures the underlying genotype if you only see the phenotype.

That makes predicting offspring blood types from parents slightly less direct than predicting Snapdragon color.

Got it.

So 282, the Narasen's heritability, is what lets us predict offspring traits from parent traits.

Yes.

There's a formula for predicting a single offspring's phenotype based on the parent's average phenotype, the mid -parent value and the population mean mu plus H2TP mu.

Let's try the IQ example.

Say population average IQ is a hundred dollars and we know 22 for IQ is around 0 .4.

Parents have IQs of 110 and 120.

Okay.

So the mid -parent value 2PT is 110 plus 120 to exoys 115.

The parental deviation for the mean is 115, $100, 15 house.

So predicted offspring IQ is $100 plus H2 times 15 phi.

That's So even though the parents are 15 points above average, the offspring is only predicted to be six points above average.

They regress towards the mean.

Precisely.

The 10, 22 of some point four tells us only 40 % of that parental superiority is expected to be passed on via additive genetics.

And this is exactly what animal and plant breeders use for artificial selection, right?

Absolutely.

They measure tour titty, sue twos for traits like milk yield or crop height.

Then they select the best individuals to be parents.

The difference between their average and the population average is the selection differential.

And the formula dollar Aquila's H2 is LR predicts the response to selection, which is how much the next generation's average will improve over the original population mean.

Exactly.

If 2H82 is high, selection is very effective and R will be large.

If 2H82 is low, even selecting the very best parents won't result in much improvement in the offspring.

It guides the whole breeding strategy.

So we've covered the quantitative traits and the how much heritability using statistics.

Now, how do we find the actual genes, the where?

This takes us into molecular territory.

We're looking for quantitative trait

loci or QTLs.

Right.

QTLs are specific regions on chromosomes that contain genes influencing a quantitative trait.

Finding them used to be incredibly laborious.

Like that tomato experiment by Steven Tangsley.

That sounds like a classic.

It really is.

He crossed a tiny wild tomato, like one gram fruit with a big cultivated one, maybe 500 grams, huge difference.

And then look to the F2 generation.

Yes.

The key was using molecular markers back then, mostly restriction fragment length polymorphisms or RFLPs.

These are like little genetic signposts scattered across the chromosomes.

So you take an F2 plant, measure its fruit weight, and also figure out which parental signposts it inherited in different chromosome regions.

Exactly.

If plants inheriting the RFLP marker from the big fruited parent consistently had heavier fruit than plants inheriting the marker from the small fruited parent, you could infer that a QTL affecting fruit weight was located near that marker.

And they found specific locations like FW2 .2 on chromosome 2.

Precisely.

They mapped several QTLs contributing to that massive difference in fruit size.

It was pioneering work showing you could dissect a complex trait down to specific genetic regions.

Nowadays, we can do this on a much grander scale, especially in humans, using genome -wide association studies or GWS.

Oh, absolutely.

GWS is QTL mapping supercharged.

Instead of a few hundred markers, you're looking at millions of single nucleotide polymorphisms, SMPs, tiny variations across the entire genome in thousands, sometimes hundreds of thousands of people.

And you're looking for SMPs that are statistically more common in people with a core idea.

It's all about statistical association.

And the results are often shown in those Manhattan plots.

Yeah, they're quite striking.

The plot shows all the chromosomes lined up on the bottom, the x -axis.

For each S &T tested, there's a dot plotted.

And its height on the y -axis represents how statistically significant the association is, usually as a negative log of the p -value.

So higher points mean stronger association.

And you get these skyscrapers sticking up, indicating regions with strong associations.

Exactly.

But because you're doing millions of tests, the threshold for significance has to be incredibly stringent to avoid false positives.

Typically, it's set at $5 x 10 .8.

Only peaks towering above that line are considered genome -wide significance.

Has GDAES actually pinpointed genes for complex human conditions?

It has.

For schizophrenia, for example, strong signals consistently pop up near the major

complex, MHC, on chromosome 6 that's involved in the immune system, and also near genes like the DRD2 dopamine receptor on chromosome 11.

It's finding real biological clues.

Okay, so we can find the genes.

Let's swing back to behavior and things like intelligence, using twin studies again.

Instead of concordance, we use correlation.

Right.

For continuous traits like IQ scores or personality measures, we calculate the correlation coefficient faulty between different types of relatives.

It measures how similar they are.

And identical twins reared apart, MZA, are particularly informative.

Extremely.

They share 100 % of their genes but grew up in different environments.

So their correlation coefficient gives a direct estimate of the broad sense heritability to H2 dollars.

Any similarity must be due to their shared genes.

And the correlations for IQ in MZA twins are, well, they're quite high, aren't they?

They are consistently high, often reported in the range of 0 .7 to 0 .8.

Wow.

So that suggests a broad sense heritability, 22 .8 or 2 dollars, for IQ scores of maybe 70 % or even higher.

That's the interpretation, yes.

It indicates a very substantial genetic influence on the variation we see in IQ test performance within the population studied.

70%.

That's a figure that always sparks debate, implying significant genetic influence on cognitive abilities.

What about personality?

Heritability estimates for major personality dimensions tend to be a bit lower, but still significant, often falling in the range of 39 % to 50 % for two H2 dollars, again, largely based on twin studies.

And even for complex behavioral issues like alcoholism.

We see evidence there, too.

Remember the concordance rates, MZ twins around 41%, DZ twins around 22%.

That difference, again, points towards a genetic contribution to the liability for developing alcoholism.

So looking back, it's quite a journey.

We started with basic observations and simple statistics, like Johansson partitioning gene E, then Fisher giving us the math for variance components, VG hours, V -day, V -day.

Now we have molecular tools like G -day ways pinpointing actual genes underlying these statistical estimates for really complex human traits.

It's a powerful synthesis of statistics and molecular biology.

We've gone from asking if genetics plays a role to estimating how much and now starting to identify which specific genes are involved.

So this leads to our final thought for you, the listener.

Given these frankly impressive heritability statistics from twin studies like that 70 % figure for IQ and the accelerating progress of G -day and finding specific genetic variants associated with intelligence, personality, and disease risk, what does this mean for society?

Yeah, as we move from just knowing the statistics to potentially knowing the specific genetic predispositions of individuals,

what kinds of shifts might we see in education, in preventative medicine, maybe even in how we assess risk or potential?

It raises some really profound questions about how we use this increasingly detailed genetic information, doesn't it?

Balancing potential benefits with ethical concerns.

Definitely food for thought.

Thank you for joining us on this deep dive into the genetics of complex traits.

We hope this exploration sparks further interest in these fascinating and rapidly evolving fields.