What percentage of the human genome needs to be error free to result in a functioning human? In other words, how many genes are there that if there’s a mere one letter error the result will be either stillbirth, severe disability, or a future incurable debilitating disease?
Well, keep in mind that every human has two full copies of their genome (with the exception of males, which only have a single X and Y chromosome). This means that for every gene, if one copy is deleted or broken by a mutation, the extra copy can fill in the same role.
Also, a mutation in a gene can range from completely harmless to completely damaging. On the harmless end, you have mutations that change the DNA in a way that doesn’t change the protein it codes for. Then you have mutations that give very minor changes that result in slightly decreased gene function. Or mutations that give bigger problems, or even mutations that completely “break” the gene.
I’ll search around to see if I can find any papers directly relevant to what you’re interested in. My first guess is that nobody’s come up with a solid estimate of the necessary genes in humans, since you can’t exactly go through and systematically knock them out. Possibly there’s been some effort in mouse models.
Sometimes, but not always. That’s why you get dominant mutations.
The answer to the OP’s question is that we don’t know. We only know the functions of less than half of our genes. We just got the genome sequenced a few years ago - it’ll take time to figure out what it all does!
However, other organisms are better understood. E. coli has about 4300 genes. Of those, 1800 are known to be lethal if knocked out. For yeast, it’s 1100 out of about 6000 genes. However, these tests were done in the lab, obviously. There are presumably a lot of genes that are not needed in a flask that would be vital out in the wild, competing for resources. Also, with bacteria and yeast, it’s hard to detect more subtle defects - what we would consider disease in humans.
Similar work was done in the worm C. elegans, and they found that about 7-10% of genes were lethal if removed, a few caused observable phenotypic changes, and ~70-80% caused no observable change at all. Again, the above caveats apply, and humans are so much more complex that I would be very surprised if our numbers were in the same range.
Now you’ve got me thinking about it, if we approach the problem from the other direction, it gets more complicated. A lot of the work going on these days is aimed at understanding how multiple genes interact to cause disease. The days of saying “we’ve found THE gene that causes X” are mostly over. Now we’re looking at situations like this: say you break gene A. No problem - you’re still fine. And say we break gene B, with the same result. But say you break both, and suddenly you’ve got some major problem. How would you classify that?
And that’s a very simple example. With the advent of microarrays and haplomaps, it’s beginning to be a matter of routine to link dozens, or hundreds, or thousands of genes to a particular condition.
If there’s one truism in biology, is that it’s more complex that you thought, no matter how complex you thought it was. That’s particularly true in human biology.
So it’s not an easy question.
Occasionally mutations are beneficial. That’s the basis of evolution.
As another complicating factor, keep in mind that some genetic illnesses don’t result from an error in DNA codons per se, but from duplication or translocation.
Down syndrome results when there are two copies of chromosome 21. Each copy may be perfectly fine, but the presence of a second copy mucks things up somehow.
Then there’s the fact that most of any organism’s genome (including ours) apparently isn’t part of any gene at all. There are vast stretches that are nothing but the same short sequence repeated over and over and over and over and over and over and over and over.
Three/third.
True, though more and more of these regions are proving to be important. Many neurodegenerative disorders are turning out to be caused by expansion of three-letter repeats within their genes. My current project is trying to figure out what some of these noncoding regions are doing in the fruit fly.
Smeghead found the numbers for model organisms before I had a chance.
One caveat on the ~7% of lethal genes in C. elegans: that number was found by a RNA interference screen. The technique is incredibly fast and easy, but pretty sloppy, so it doesn’t always knock out the gene as much as you want or even at all. Basically, the 7% figure is a good lower bound, but probably a significant underestimate.
Thanks for the answers. Do we know of any major congenital defects that are the result of an error in one nucleotide?
Cystic fibrosis is caused by a mutation in one gene, but you usually need both copies to be mutated, and the most common mutation is a deletion of three nucleotides. I’d have to think more on that. I could list off several other diseases that are caused by one nucleotide mutations, but I wouldn’t call them major congenital defects.
Another complication is epigenetics, changes that can be passed from generation to generation with no change in DNA such as which genes are turned on or off.
Two people could have the same DNA but one healthy while the other very sick because of which genes were expressed.
Yes, but many of those can be traced back to mutations. Prader-Willi/Angelman syndromes are the classic example of a disease caused by a mistake in methylation patterns, but the reason the patterns are off is because of a deletion in the DNA.
Looking back at the OP, the question was whether a single change in one base can make an organism non-viable and the answer is pretty clearly no. First off, many such changes have no effect whatever since often 2 or 4 three letter bases lead to the same amino acid. Then in any protein, there is usually one important region and the rest is scaffolding and most changes in one amino acid will make little difference.
That’s not correct. A single nucleotide change definitely can completely knockout a gene. First, you can have a frameshift mutation, which often result from a single nucleotide deletion or addition. These usually result in a completely useless protein, though if it happens near the end and misses the important domains the protein still can retain some function. Second, you can have a nonsense mutation, which gives a premature stop codon. This “deletes” everything after that point for the protein product, which usually is a pretty big problem, unless it just trims a bit off of the end. Then there are missense mutations, which substitute one amino acid for another. Most of the time, these have little (if any) effect, since redundancy in the genetic code gives a higher probability of substituting a similar amino acid. Still, there can be changes in a critical domain, which again can damage or remove the protein function. And then there are some less straightforward cases, where a mutation changes mRNA splicing, or regulatory elements.
Now, most of the time, single nucleotide changes are silent (giving no change to the protein encoded). Most of the remaining mutations are very minor. Only a small subset will really break the protein encoded. Still, they do exist, and are responsible for a number of human diseases.
Examples of diseases from these kinds of mutations include Tay-Sachs, cystic fibrosis, some forms of muscular dystrophy, sickle cell anemia, and spinal muscular atrophy. We don’t have good examples in humans for mutations that are lethal before birth, because nobody studies the genetic causes here. There are a lot of stillbirths, spontaneous abortions, nonviable embryos, etc., and at least some of these are due to genetic deficiencies. It’s estimated that each human caries about 5 recessive lethal mutations, on average. These don’t show up unless someone reproduces with another person that happens to carry one of those mutations, which is rare except in cases of inbreeding. These recessives are also probably at least partly responsible for low fertility associated with lots of inbreeding.
There are no absolutes in biology. I guarantee it. 
If you can imagine it, if you look hard enough, you’ll find an example. I’m sure somewhere out there there’s one nucleotide that can’t be changed without major problems.
The OP asked if there was DNA so sensitive that a single change in a codon would necessarily render an organism non-viable. And that is what I answered. Of course there are places in which a single change could have that effect. And it was clear that the OP was about a SNP, not any more profound change like an omitted or duplicated region.
But the examples lazybratsche mentioned are SNPs and also “future incurable debilitating diseases”
I was thinking of not just non-viable, but any genetic defect that would require major post-natal surgery such as a seriously defective heart, a lifetime of hospitalized care, a complete lack of one of the senses, severe reactions to everyday stimuli such as sunlight, or a debilitating incurable disease anytime in the persons life, such as Tay-Sachs or Huntington’s. In short, any defect that would cause the person to be fated to die early, or live miserably.