Re: [DIYbio] Re: Anand - Lets FFT that wet stuff ...

Thanks for the Document URLs .... This one seemed comprehensible to me, though all of them are fairly straightforward really.

https://www.researchgate.net/publication/5260702_Accelerating_and_Focusing_Protein-Protein_Docking_Correlations_Using_Multi-Dimensional_Rotational_FFT_Generating_Functions


TV MPEG-4 compression uses a similar technique to compress video, it breaks a field of vision into zones, each has a datum in the middle and sine and cosine wave ripples out from the centroid. a delta from the two matches zones to see which way there moving. Then, often the update is just a X Y new datum. Its like 20:1 compression and is slightly lossey.

Alphafold-2 discards absolute positioning of proteins entirely early on, all delta from features are used so there is no magic 'best' facing angle. The last steps to undo that is full of nasty complex tradeoffs.

For investigations of most life science questions, M.D. for small subsets of molecules is probably always worth doing, for instance before even speculating on wet chemistry. So perhaps most frequently protein-B.P. investigations are as much a process of rapidly discarding no-go subsets... ?

When any things 'fit' the conjugate in some enumeration are perfect fits, like i + j stuff. Simplifying investigations to pairs is pretty dangerous to deep comprehension. I don't know of a matching algo for greater then pairs, but there is an algo called Minikowsi distances I think might apply. What I suggest is a context sensitive detection perhaps is best for discarding complete no-go's and make candidate ones queue up for M.D. once identified.  I think if the slope of the minikowski derived tray is flat, the matches in whatever context make the independant variable vectors are good fits.

It seems like the papers take various aspects and leap into comparisons with previous benchmarks without any exposed speculation on why such a thing should work. 

If you decide to try to implement a FFT in a chipset optimized for such a thing, let me know, Maybe I can throw iin on some aspect of that specifically. I don't quite 'get' why yhr trig radiating out of the centroid of an image is better then other numerical methods either. But the compression is done in r/t without unusual semiconductors... Just chips with a few specialised instructions here and there.

Maybe an interactive IDE might help the usability of these programs. Like the last steps in alphafold for shape determination going from a position in a list to a rendering is a fussy step. 

Regards,
Daniel B. Kolis
 

--
-- You received this message because you are subscribed to the Google Groups DIYbio group. To post to this group, send email to diybio@googlegroups.com. To unsubscribe from this group, send email to diybio+unsubscribe@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/diybio?hl=en
Learn more at www.diybio.org
---
You received this message because you are subscribed to the Google Groups "DIYbio" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diybio+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/diybio/6f383108-408c-4034-adca-ae54319155a5n%40googlegroups.com.

  • Digg
  • Del.icio.us
  • StumbleUpon
  • Reddit
  • RSS

0 comments:

Post a Comment