More Recent Comments

Sunday, July 22, 2007

The Frequency of Alternative Splicing

One of these days I'm going to get around to blogging about alternative splicing. As most of you know, the databases are full of information about alternatively spliced gene products in mammalian genomes. There are many scientists who believe that most mammalian genes have two or more different products as a result of alternative splicing of the primary transcript.

I think it's nonsense. When I look at my favorite genes, the HSP70 gene family, the predicted protein products make no sense whatsoever. (The image shows predicted splice variants of the HSPA5 (BiP) gene from the SpliceInfo database.) The alternatively spliced variant often removes a piece of the hydrophobic core of the protein or other parts that are known to be essential. Since these proteins are the most highly conserved proteins in all of biology, it makes no sense at all to predict that mammals have all of a sudden evolved variants that are missing large hunks of highly conserved amino acid sequence. I think that most predicted splice variants are artifacts of the EST databases.

The annotators of the human genome have pretty much rejected all of the splice variants of HSP70 genes (e.g., Entrez Gene HSPA5) and many other genes whose structures are known. They have not rejected the multiple splice variants of other genes that are less well studied.

Anyway, like I said, this discussion will have to wait for another time. Meanwhile, you can read my friend Deanne Taylor's views on alternative splicing (and her disagreement with me) on her blog [Alternative Culture]. She will be here in Toronto next summer to debate the issue so everyone should plan on attending a mini-Howlerfest. It will be at the same time as the Darwin Exhibit at the ROM [Charles Darwin Is Coming to Toronto].


RPM said...

It seems like this could be resolved by seeing if alternate splicoforms are conserved across multiple taxa. For example, we can ask if a putatively alternatively spliced gene in the human genome only has one splice variant conserved amongst primates or mammals. If we see that most alternative splice variants are not conserved, then that would suggest that they are just noise. If, on the other hand, we see that they are conserved, then that would suggest that they are functional. Of course, that assumes that there is not rapid turnover of alternative splice variants.

Larry Moran said...

They are not conserved. The people who are pro alternative splicing take this to mean rapid evolution among mammals. The fact that humans have different splice variants than mice is supposed to be one of the main things that make humans different than mice.

I kid you not.

Peter said...

Hmm, have the splice variants been double-checked (e.g. by RT-PCR) to see that they really do exist as transcripts? Surely that has to be the absolute first step?

After that, you can move on to the question of whether it produces a protein at all.

Larry Moran said...

Peter asks,

Hmm, have the splice variants been double-checked (e.g. by RT-PCR) to see that they really do exist as transcripts? Surely that has to be the absolute first step?

Yes, in some cases labs have tried very hard to detect these unusual transcripts but they've failed to find them. Of course this data can't be published because it's a negative result.

The proponents of frequent alternative splicing rely on large-scale analyses of entire genomes to support their case. It's easy for them to dismiss the results for a single gene.

The skeptics are not about to waste time refuting each and every predicted alternative transcript for dozens of genes.

In an ideal world, the shoe should be on the other foot. Those who claim that these transcripts are functional should have been required to offer evidence that these weird transcripts are biologically significant. That didn't happen. Instead, the idea that every human gene could produce multiple proteins crept into the literature in spite of the fact that it goes against everything we know about evolution and gene expression.

That's the strange thing about science these days. We're seeing a great many claims like this becoming widely believed not because of solid evidence but because it fits with certain pre-conceived notions. In this case, many people were upset that we had only 25,000 genes and those genes seemed to be very similar to the ones in a rat. The idea of frequent alternative splicing appealed to them because it allowed for a proteome of several hundred thousand proteins and it "explained" how humans could be so different from rats even though their genes seemed to be similar. (Humans have evolved different proteins by alternative splicing.)

SPARC said...

Thanks for bringing up this issue. Have you tried to find a constitutively spliced gene in ASD? Even those genes familiar to me and those which have been published as being not alternatively spliced in the 90s are displayed with several alternative exons. IMO alternative splicing databases are nature's revenge for EST sequencing and for the overoptimistic omics views on biology.

Anonymous said...

I'm a bit confused - are we arguing that alternative splicing doesn't occur on the scale that is seen in databases, or that it isn't functionally significant?

These are two different issues. It's quite conceivable to me that widespread alternative splicing is not only real, but inevitable. And that many or most splice variants actually have no "function".

Larry Moran said...

Art said...

I'm a bit confused - are we arguing that alternative splicing doesn't occur on the scale that is seen in databases, or that it isn't functionally significant?

I'm arguing that it doesn't occur on the scale that is claimed. In my opinion, the conclusion that an alternatively spliced message exists is often based on artifacts. The most common artifact is a so-called "expressed sequence tag" (EST) that is really a splicing error.

Anonymous said...

Thanks Larry.

If I understand you correctly, your statement leads to a follow-up question - namely, what is the difference between a "splicing error" and an alternative splicing event?

Larry Moran said...

A splicing error is something that occurs infrequently and unpredictably when the primary transcript fails to be processed in the normal way. You distinguish splicing errors from normal alternative splicing events because the errors cannot be reproduced and the product has no function.

The error rate in splicing appears to be quite high. This is probably because the 5' and 3' splice site consensus sequences are so short.