A lot of noise has been made about the probability of finding functional sequences at random. So, it was interesting to stumble across a paper that tests how hard that is. The researchers basically hooked up a bacterial sequence that ensures neighboring sequences are made into RNA, and the randomized the neighboring sequences. They then exposed the bacteria to antibiotics and checked for resistance. Overall, they tested 100 million random sequences, which is well below some of the probabilities said to be needed for function in other contexts.
They came up with six different sequences that conveyed antibiotic resistance. They identified a bacterial gene that's required for all six to work. As near as they can tell, all six bind to the protein made by this gene, altering its activity. The gene is responsible for some of the interactions between the bacterial cell wall and cell membrane, and the antibiotic alters the cell membrane, so it all seems to fit together.
All six of the new proteins appear to sit inside the cell membrane. Half of them are small enough (~25 amino acids) to be nearly entirely within the membrane. The rest are closer to 50 amino acids, and probably pass through the membrane, loop around, and pass through it again.
A couple of thoughts on what this result says about probabilities in general:
The six different proteins don't have a lot in common other than being able to sit within a membrane - the sequences aren't especially related. Which means there's probably a very large number of sequences beyond the ones identified here that would be functional. This is why probability arguments that focus on having one specific gene sequence are stupid - we don't know how many different sequences can possibly perform the same function.
The second thing is that we have no idea how many of the remaining 100 million sequences that didn't work here actually do perform functions that we haven't tested for. Because the probabilities are all about doing something useful, not performing one specific function.
The argument could be made that these only work because other proteins are already around and doing important things. Which is true. But that's also been true since the moment the first self-replicating RNA existed - there has always been something around doing important things since life existed. So, while the argument could be made, it probably shouldn't be.
In any case, the paper:
https://journals.plos.org/plosgeneti...l.pgen.1009227
They came up with six different sequences that conveyed antibiotic resistance. They identified a bacterial gene that's required for all six to work. As near as they can tell, all six bind to the protein made by this gene, altering its activity. The gene is responsible for some of the interactions between the bacterial cell wall and cell membrane, and the antibiotic alters the cell membrane, so it all seems to fit together.
All six of the new proteins appear to sit inside the cell membrane. Half of them are small enough (~25 amino acids) to be nearly entirely within the membrane. The rest are closer to 50 amino acids, and probably pass through the membrane, loop around, and pass through it again.
A couple of thoughts on what this result says about probabilities in general:
The six different proteins don't have a lot in common other than being able to sit within a membrane - the sequences aren't especially related. Which means there's probably a very large number of sequences beyond the ones identified here that would be functional. This is why probability arguments that focus on having one specific gene sequence are stupid - we don't know how many different sequences can possibly perform the same function.
The second thing is that we have no idea how many of the remaining 100 million sequences that didn't work here actually do perform functions that we haven't tested for. Because the probabilities are all about doing something useful, not performing one specific function.
The argument could be made that these only work because other proteins are already around and doing important things. Which is true. But that's also been true since the moment the first self-replicating RNA existed - there has always been something around doing important things since life existed. So, while the argument could be made, it probably shouldn't be.
In any case, the paper:
https://journals.plos.org/plosgeneti...l.pgen.1009227
Comment