Re: Orr on No Free Lunch

From: bivalve (bivalve@mail.davidson.alumlink.com)
Date: Mon Sep 02 2002 - 14:56:23 EDT

  • Next message: John W Burgeson: "Re: New society formed"

    DC: Both of us are agreeing that much is unknown; we differ in our
    opinions as to what seems most likely regarding this unknown. It's a
    bit like arguing if the glass is half empty or half full when we can
    only see a tiny fraction of the top and bottom of the glass.

    >>DC: First, although there may be no guarantee of finding a free
    >>lunch, free snacks (aka local maxima) are easy to find. A simple
    >>hill-climbing algorithm will achieve this. A slightly more complex
    >>algorithm, which gives some probability of descending as well as
    >>ascending, will be less likely to get trapped in a small local
    >>maximum.

    >JB--How do computer algorithms relate in any way to functional
    >protein sequences. How does the computer decide when a given
    >protein sequence will be soluble, fold or have any function
    >whatsoever? How do the concepts I presented from protein folding
    >issues incorporate into the algorithm. A local maximum found by an
    >algorithm, when translated into a protein, yields a non-functional,
    >non-folding product. Until these issues are addressed by the
    >algorithm, they are clever math projects, nothing else. <

    DC: Dembski uses genetic algorithms as examples of the No Free Lunch
    principle. They also represent the quickest way to model an
    evolutionary process. Precise matching between the algorithm and
    proteins is impossible given our current inability to predict details
    of protein structure and function. In vitro searches for functional
    genes or experimental evolution of bacteria and viruses can provide
    actual examples of protein evolution, but are much slower. Also, I
    use algorithms but work on sequencing existing genes rather than
    evolving new ones, so I am familiar with what certain algorithms can
    do but not with details of experimental protein evolution.

    >>DC: This second, more complex algorithm seems like a good analogue
    >>to the combined effects of genetic drift and natural selection. <<

    >JB: --Which assumes fully functional biological systems varying
    >across large populations. It has nothing to do with protein
    >biochemistry. <

    DC: As these functional biological systems use a lot of protein
    biochemistry, there is a connection. Mutation can produce novel
    proteins, which may or may not be any good. Genetic drift and
    selection determine whether these proteins spread in the population.

    >>DC: A mutation has a probability of spreading in a population that
    >>is a function of the merit of the mutation (as well as population
    >>size, initial frequency of the mutation, etc.) Unless the mutation
    >>is significantly detrimental, there is some probability of its
    >>persisting, allowing evolution to traverse suboptimal valleys to
    >>reach more optimal maxima. If a substantial proportion of the
    >>possible mutations are highly detrimental, this may make it easier
    >>to find good options because so many options are removed from
    >>consideration. <<

    >JB: --Conversely it would take many more generations of species
    >reproducing incorporating mutations at the given random rate than
    >have ever existed and therefore this sampling would not be possible.
    >You would basically never find the good options once the negative
    >ones were found because it would take much more time to go forward
    >than to go backward. <

    DC: This depends on the starting point, the goal, and the topology of
    the adaptive landscape. The sampling is quite possible in some
    cases, but in general the necessary sample size is unknown,
    especially in light of the possibility of multiple paths and multiple
    adequate local maxima. The establishment of good options actually is
    easier once bad ones are found because good is relative. The
    selective pressure in favor of a better option is greater if the
    option is much better than if it is only a little better. However,
    major deletions (usually detrimental) are difficult to reverse. At
    the same time, such major disruptions often prevent survival or
    reproduction and thus never become established.

    >>DC: I presume you mean that this class has a distinctive
    >>three-dimensional structure and function not found in other
    >>classes. But how distinctive is distinctive? Some basic
    >>structural features are shared by thousands of proteins,..... <<

    >JB: --I'm guessing if we were both X-Ray crystollagraphers there
    >would be no confusion on this. My understanding is that while
    >proteins in general have primary (sequence), secondary (alpha
    >helices, Beta strands, etc.), tertiary (global fold) and quaternary
    >structure elements, they are very unique. The secondary structural
    >elements are common to all proteins, however these secondary
    >structural elements do not confer function-- only the fully folded
    >global/quaternary structure enables the protein to function. A
    >sequence that has elements that can form several secondary
    >homologies like an alpha helix or beta strand, but lacks the ability
    >to maintain any tertiary fold will not function in a way that
    >confers natural selective advantage. Thus sequences that bear close
    >relationship to the final sequence look close by their primary
    >structure, but can be quite different in the higher order
    >structures. Many residues that can be flexible in amino acid
    >changes cannot be changed fr!
    om say positively charged residue to hydrophobic, because this would
    have great influence on the tertiary and quaternary structure, making
    the final product unrecognizable structurally. Thus the previous
    statement from Stryer Biochemistry text highlights how the changes in
    primary sequence have huge affects on the tertiary sequence and
    delineates how constricted the changes from primary sequence are
    before you lose functional tertiary and quaternary structures. Which
    leads to your statement: <
    >>DC: whereas a change of a single amino acid is a recognizable
    >>difference, though perhaps without impacting the folding or
    >>function in any way. <<

    >JB: --Thus to speculate on "perhaps," we need to be much more
    >informed on the changes that are tolerated in any given protein
    >sequence. Computer algorithms completely ignore this essential
    >issue of protein biochemistry/ folding in my experience. <

    DC: However, this cuts both ways. A single amino acid change may
    ruin a protein, but it may also turn it into something with a novel
    useful function. The more tightly you constrain the definition of
    distinctive kinds of proteins, the fewer changes are needed to evolve
    from one kind to another; the more broadly you define kinds, the
    fewer kinds there are that must be found by evolution. We do have
    some information on the amount of change tolerable in protein
    sequences based on the range of functional sequences found in living
    things. However, this is only a subset of the possible functional
    sequences, not to mention the numerous living things for which we
    have little or no data.

         Dr. David Campbell
         Old Seashells
         University of Alabama
         Biodiversity & Systematics
         Dept. Biological Sciences
         Box 870345
         Tuscaloosa, AL 35487 USA
         bivalve@mail.davidson.alumlink.com

    That is Uncle Joe, taken in the masonic regalia of a Grand Exalted
    Periwinkle of the Mystic Order of Whelks-P.G. Wodehouse, Romance at
    Droitgate Spa



    This archive was generated by hypermail 2.1.4 : Mon Sep 02 2002 - 15:15:50 EDT