We looked at BLOSUM62 scores and found they gave similar results - the very closet proteins are tagRFP derivatives according to both metrics. The outlier dot below is the circularly permuted cp-mKate which BLOSUM scores higher, but still in consistent range with closest FPs

Another really interesting question! Based on your tweet I put together this quick analysis this morning (sample size small!) using natural FP's but excluding sequences from the family they come from

I think its an interesting question - we were actually inspired by your last tweet to compute this metric too! Great to see we get the same 74% number for two sequences, this is what we see for 3, 4, 5 ... etc greedy coverage

We performed an evolutionary analysis of esmGFP using information from natural GFPs found in corals and sea anemone and estimate that esmGFP represents an equivalent of over 500 million years of evolution from the closest protein that has been found in nature!

C10, which we named esmGFP, has 96 edits throughout its sequence and structure from tagRFP, its closest known protein. Visualizing these edits on the ESM3 structure prediction of esmGFP we can see 22 edits occur in the protein’s highly sensitive interior

In the first plate (left) we were intrigued by B8. While 50x dimmer than natural GFPs, 43% of its sequence differs from the closest natural protein! Continuing the chain of thought from B8 on the second plate (right), ESM3 found C10, which is similarly bright to natural GFPs!!

We prompted ESM3 through a chain-of-thought leveraging existing structural and evolutionary knowledge of what might make for a successful fluorescent protein, but without fine-tuning on high-throughput experimental fluorescence measurements

Flamingo's also punching above their weight - are they super smart, but also just super chill 🦩? Also, poor cats 😹 Note I made this graph in ~2 minutes with ChatGTP4o, spot checked some numbers, seemed reasonable. Was pretty fun!!

End of an era! Yesterday was my last day @cziscience. Many thanks to all my former colleagues and collaborators for a great six years!! I'm now onto my next adventure focused on AI x Bio. Stay tuned ... 🚀

i love these answers though