Thread by @ishikaluthra

How much of the stuff that our genomes do is actually functional? We set out to answer this long debated question by creating a null hypothesis in yeast and human in the de Boer Lab's latest preprint: 1/ biorxiv.org/content/10.110…

It's often assumed that any reproducible activity of the genome (eg histone modification, transcription) is functional; but is this really true? Without knowing what to expect by chance (e.g. in the absence of selective pressures), this is difficult to answer. 2/

To explore this question in yeast, we transcriptionally profiled ~760kb of human DNA in yeast and because humans and yeast are separated by ~1 billion years of evolution this DNA is "evolutionarily naive" to yeast. 3/

Shockingly, we found the ~760kb of human DNA to be extensivley transcribed in yeast! The transcripts mimicked the length and expression level of the evolved yeast genome, and the promoters used in yeast were distinct from human promoters. 4/

Adding swaths of naive DNA is much trickier in human cells. Instead of directly testing naive DNA in human cells, we did the next best thing: predict activity with a state of the art sequence-to-gene regulation model called Enformer (Avsec et al. Nat. Methods. 2021). 5/

We also saw lots of activity in naive DNA in humans too, but it requires that you preserve local dinucleotide content. 6/

People have used markers like cell type specificity and co-occurance of chromatin marks as indicators of selections. After all, why would an organism bother to express something in a cell-type specific way with multiple chromatin marks if it were unimportant? 7/

Instead, we find that naive DNA is more cell type specific in it's gene expression than the evolved genome, suggesting that cell type specificity may not be a reliable indicator of function. 8/

Further, we find similar patterns of co-occurance of chromatin marks in naive and evolved DNA, again suggesting that using co-ocurance of chromatin marks as evidence of functionality may not be the best approach. 9/

However, we did find that in both yeast and human, the extremes of expression/chromatin marks do appear to be reliable indicators of selection and are much rarer in naive DNA. 10/

In conclusion, our first-pass functional genomics null hypothesis showed that we expect a lot of gene regulatory activity by chance... and maybe, just maybe, not everything that the genome does is functional. 11/

As the scale of DNA synthesis improves, improving this null with empirical data will be critical. 12/

Thanks all the other authors of this paper for all their help with this project! @CarldeBoerPhD @EmiliaCXY @CassPJensen @muntakim_rafi @asfarlathif. 13/

Surprisingly, many research groups seem to have had similar ideas simulataneously, including these two recent preprints that you should check out if you enjoyed this thread: 14/
biorxiv.org/content/10.110…
biorxiv.org/content/10.110…