I just checked my profile, and it seems that a post I shared a few days ago has also led to the account being deleted.

This is not proof of anything, and I can't exactly remember what was the original post that was deleted, or the account, but it seems to corroborate the other reports.

Anyone want to listen to music from protein sequences (technically NMR data)?

I was wondering when I would get round to integrating my protein and music evolution research themes, but it looks like someone else has got there first!

www.biorxiv.org/content/10.1...

I can barely believe the things I've seen of the Duolingo owl

5/5 We show using a simple model of melody generation, that empirical constraints (melody length, scalar motion, and primarily information rate) are sufficient to reproduce the number of notes in scales (a well-documented universal feature of melodies).

3/5 Eurasian music-theoretic scales almost perfectly fit harmony theory, which explains the dominance of harmony theory amongst Western scholars. We speculate that mathematical development and tuning technology is what drove the evolution of music-theoretic scales.

1/5 Preprint alert – horn tootin time!
Proud to present two megaprojects: the end product of 6 years of work on melodies and scales. One comparison of theories of the evolution of scales using global scale data. One on cross-cultural constraints on melodies and scale size.
#musicscience#musicpsych

Really excited to be back working on the evolution of musical scales. My project is coming to an end at last after 6 years.
🎵🧬 🎵🧠

First, I created a database of scales from data in the ethomusicology literature. Kudos to Ellis, Kunst, Tracey, Kubik, and all who followed in their footsteps.
1/5

Amazing, you'll do a great job. They probably want to know what to put on the next golden record they send out.

Meanwhile, here's an e-mail I got yesterday. I really appreciated the 90s web aesthetics.

I find that most mutations that I've seen (in the PDB, so some survivor bias there) do not lead to structural changes that are easily observable. E.g. here's a snapshot of one of the largest deformations from a single substitution (V66K).

For small effects, using many AF structures helps.
🧶🧬

Another advance is that we combine multiple AF-predictions by averaging local structures. This greatly reduces noise due to stochastic fluctuations in flexible regions.

We released a python package for these methods, PDAnalysis
github.com/mirabdi/PDAn...
5/5

One key advance is the development and testing of different metrics of structural change. Summary: LDDT is bad (low res.); we use effective strain (like FAPE, but normalized by distance). Previous analyses with opposite conclusions failed because they used RMSD and ΔpLDDT.
4/5

AF structures can be used to predict changes in fluorescence (8000 BFP variants). Unlike the previous analysis, most of these variants are not in the PDB. This means AF doesn't work simply by memorizing structures.
3/5

Evaluating AF using PDB data is difficult because we can't distinguish mutation effects from stochastic noise - we can't say whether individual predictions are correct. Instead, we show that correlations between PDB-AF structures are higher when mutations are present than if they are not.
2/5

We have an upcoming pre-print that shows exactly this! We estimated deformation due to mutation using AlphaFold and found that it correlates quite well with changes in stability. Large effects on structure appear to lead to large effects on stability.