Recipe: Local Mutate 3.0

created by spvincent

Profile

• Name: Local Mutate 3.0
• ID: 103229
• Shared with: Public
• Parent: None
• Children:
  • Local Mutate 3.0 w/bands
  • Local Mutate 3.1.1
  • Mutate Mutate 3.0
  • Mutate Local 3.0
  • Local Mutate 3.2
  • Freeze sidechains
• Created on: January 13, 2020 at 21:36 PM UTC
• Updated on: January 13, 2020 at 21:36 PM UTC
• Description:

  See first comment

Code

min_residue = 1
max_residue =  999
n_residues = 0
n_mutations = 0

delta_residue = 9
kMaxDelta = 20

best_score = 0
clash_importance_during_modify = 0.09

kWiggleIterations = 15 

modify_count = 10
original_slow_filters_setting = 0

score_type = 1

loss_from_modify = {}

original_residues = {}
new_residues = {}

scores = {}
ids = {}

start_time = 0

wiggle_sidechains_first = false

--    Amino acids. Hydrophobics first.

single_letter_codes = { "g","a","v","l","i","m","f","w","p","s","t","c","y","n","q","d","e","k","r","h"}
three_letter_codes  = { "Gly","Ala","Val","Leu","Ile","Met","Phe","Trp","Pro","Ser","Thr","Cys","Tyr","Asn","Gln","Asp","Glu","Lys","Arg","His"}

        -- Use of quicksave slots

kOriginalStructureOrNewBest = 10
base_primary_sequence = {}
n_sequence_changes = 0

function r3 ( x )

-- Round to 3 decimal places

    t  = 10 ^ 3
    return math.floor ( x*t + 0.5 ) /  t 
end

function GetScore ()

    if ( score_type == 1 ) then

        score = current.GetScore () 

    elseif ( score_type == 2 ) then

        behavior.SetFiltersDisabled ( false )
        score = current.GetScore () 
        behavior.SetFiltersDisabled ( true )

    end

    return score

end

function aa1_to_aa3 ( k )

  for  j = 1, 20 do
        if ( k == single_letter_codes [  j ] ) then
            return three_letter_codes [  j  ]
        end
    end

    return "Unk"

end

function get_aa3 ( i )

    k = structure.GetAminoAcid (  i )

    for  j = 1, 20 do
        if ( k == single_letter_codes [  j ] ) then
            return three_letter_codes [  j  ]
        end
    end

    return "Unk"

end

function GetBasePrimarySequence ()

    for i = 1, n_residues do
        base_primary_sequence [ i ] = structure.GetAminoAcid ( i )
    end

end

function GetNumberOfPrimarySequenceChanges ()

    n_changes = 0
    for i = 1, n_residues do
        if ( base_primary_sequence [ i ] ~= structure.GetAminoAcid ( i ) ) then
            n_changes = n_changes + 1
        end
    end

    return n_changes
end


function ShellSort ()

-- Adapted from Numerical Recipes in C

    inc =  1
    
    repeat 
        inc = inc * 3 + 1
    until  inc > n_residues

    repeat

        inc = inc  / 3 
        inc = inc - inc % 1

        for i = inc + 1 ,  n_residues do

            v =  scores [  i  ]
            w = ids [ i ]

            j = i

            flag = false 

            while (  flag == false and scores [  j - inc ] > v ) do
                scores [  j  ] = scores [  j  - inc ]
                ids [ j ] = ids [ j - inc ]

                j = j - inc
                
                if ( j <= inc ) then
                    flag = true
                end

            end

            scores [ j ]  = v
            ids [ j ] = w
        end

    until  inc  <= 1

end


function PostModify ( i )

    behavior.SetClashImportance ( 1.0 )

    selection.SelectAll ()
    if ( wiggle_sidechains_first == true ) then
        structure.WiggleSelected ( 1 , false , true )    -- sidechains
    end
 
    score_before_wiggle = GetScore ()
   
    structure.WiggleSelected ( kWiggleIterations )        
    score_after_wiggle = GetScore ()

end


function MutateResidue ( i  )

    local	j
    local 	k

    save.Quickload ( kOriginalStructureOrNewBest )

     for  j = 1 , n_residues do
        scores [ j ] = structure.GetDistance ( i , j )
        ids [ j ] = j
    end 

    ShellSort ()

    selection.DeselectAll ()
    for j = 1 , modify_count do
        selection.Select ( ids [ j ] )
    end

    behavior.SetClashImportance ( clash_importance_during_modify )

    structure.MutateSidechainsSelected ( 1 )  
 
    score = GetScore ()
    table.insert ( loss_from_modify , best_score - score )

    n_mutations = GetNumberOfPrimarySequenceChanges ()

 --   print ( "Hid " .. r3 ( scores [ modify_count ] ) .. " lim " .. r3 ( best_score - score ) .. "  nc " .. n_mutations )
    print (  "      lom " .. r3 ( best_score - score ) .. "  nc " .. n_mutations )


    for k = 1, n_residues do
        if ( base_primary_sequence [ k ] ~= structure.GetAminoAcid ( k ) ) then
            print ( "         " .. k .. "  "  .. aa1_to_aa3 (  base_primary_sequence [ k ] ) .. " -> " .. aa1_to_aa3 ( structure.GetAminoAcid ( k ) ) )
        end
    end


    if ( math.abs ( score - best_score ) > 0.5 ) then
        PostModify ( i )

        score = GetScore ()
        if ( score > best_score ) then

            save.Quicksave ( kOriginalStructureOrNewBest )
            GetBasePrimarySequence ()
            best_score = score
           print ( "Improvement to " .. r3 ( best_score ) )
       end

    end

end

function MutateAll ()

    n = 1

    for start_idx = 1 , delta_residue do

        i = min_residue - 1 + start_idx
        
        while ( i <= max_residue ) do

            print ( get_aa3 (i) .. " " .. i .. " (" .. n .. "/" .. max_residue - min_residue + 1 .. ")" )

            MutateResidue ( i )
            i = i + delta_residue
            n = n + 1
        end

    end

end

function GetParameters ()

    local dlog = dialog.CreateDialog ( "Local Mutate 3.0" )

    dlog.min_residue = dialog.AddSlider ( "Min residue" , 1 , 1 , n_residues , 0 )  
    dlog.max_residue = dialog.AddSlider ( "Max residue" , n_residues , 1 , n_residues , 0 )  
    dlog.delta_residue = dialog.AddSlider ( "Delta residue" , delta_residue  , 1 , kMaxDelta , 0 )  
    dlog.cidm = dialog.AddSlider ( "Clash importance" , clash_importance_during_modify , 0 , 1.0 , 2 )  
    dlog.modify_count = dialog.AddSlider ( "Modify count" , modify_count , 1 , 40 , 0 )
   
    dlog.score_type = dialog.AddSlider ( "Score type" , score_type , 1 , 2 , 0 )  
    dlog.tp2 = dialog.AddLabel ( "1 = Normal : 2 = Normal for Filters"  )

    dlog.ok = dialog.AddButton ( "OK" , 1 )
    dlog.cancel = dialog.AddButton ( "Cancel" , 0 )

    if ( dialog.Show ( dlog ) > 0 ) then
    
        min_residue = dlog.min_residue.value
        max_residue = dlog.max_residue.value
        delta_residue = dlog.delta_residue.value
        clash_importance_during_modify = dlog.cidm.value
        modify_count = dlog.modify_count.value
        score_type = dlog.score_type.value

        return true
      else
        return false
    end

end

function main ()

    band.DisableAll ()
    n_residues = structure.GetCount ()
 
    save.Quicksave ( kOriginalStructureOrNewBest )
    GetBasePrimarySequence ()

    original_slow_filters_setting = behavior.GetSlowFiltersDisabled ()

    if ( GetParameters () == false ) then
         return                -- graceful exit
    end

    print ( "Local Mutate 3.0" )
 
    max_residue = math.min ( n_residues , max_residue )
    min_residue = math.max ( 1 , min_residue )

    print  ( "Mutate range " .. min_residue .. " to " .. max_residue )
    if ( max_residue < min_residue ) then
        print ( "Mutate range error." )
        return
    end
 
    print  ( "Delta " .. delta_residue )
    print  ( "CI on modify " .. r3 ( clash_importance_during_modify ) )
    print  ( "Modify count " .. modify_count )
    print  ( "Score type " .. score_type  )
    print  ( "" )

    best_score = GetScore ()
    print (  "Start  score : " .. r3 ( best_score ) )

    for i = 1 , n_residues do
        original_residues [ i ] = structure.GetAminoAcid ( i )
    end

    start_time = os.time ()

    MutateAll ()

    cleanup ()

end

function cleanup ()
   
    print ( "Cleaning up" )
    behavior.SetClashImportance ( 1.0 )
    save.Quickload ( kOriginalStructureOrNewBest )

    selection.SelectAll ()
    band.EnableAll ()

    behavior.SetSlowFiltersDisabled ( original_slow_filters_setting )

    if ( #loss_from_modify > 1 ) then
        table.sort ( loss_from_modify )
        mid_pt = math.ceil ( #loss_from_modify / 2 )
        print ( "Median loss from modify " .. r3 ( loss_from_modify [ mid_pt ] ) )
     end

    for i = 1 , n_residues do
        new_residues [ i ] = structure.GetAminoAcid ( i )
    end

    for i = 1 , n_residues do
        if ( new_residues [ i ] ~= original_residues [ i ] ) then
            print ( i .. " : " .. aa1_to_aa3 ( original_residues  [ i ]  ) .. " -> " .. aa1_to_aa3 ( new_residues  [ i ] ) )
        end
    end

    end_time = os.time ()
    print ( "Elapsed time : " .. os.difftime ( end_time , start_time ) .. " secs" )

end

--main ()
xpcall ( main , cleanup )

Comments

[spvincent]

Fairly minor update, changing a few parameters so that the defaults work well under most circumstances. I've found this a most useful midgame script.

A script similar in style to Rav Mutate, intended for design puzzles.

Algorithm outline

For each amino acid
Set the Clash Importance to something low
Select everything in the immediate neighbourhood of that residue
Mutate
Set the CI to 1 and select the whole protein
Wiggle
Record any improvement
end

Dialog settings

Min residue, Max residue

The range to be mutated : by default the whole protein.

Delta residue

Processing each residue in turn is inefficient: the selection in each case will be largely duplicated. This setting indicates the difference between successive residues. For example, if the min and max residues above are 1 and 100 respectively and the delta value is 5, then residues will be processed in the order 1,5,10,15,etc…..2,6,11 and so on.

Clash importance

The value during the initial mutate. Something that reduces the energy score by about 200 seems about right (the median reduction is printed out at the end of the script)

Modify count

Determines how many amino acids will be mutated at each step. They will be the closest (as measured by the distance between alpha carbons) to the residue being processed. Fewer means faster processi8ng.

Score type

If set to 2, filters are toggled on to determine the score only: other operations such as wiggle are done with filters off. If set to 1 (default), the initial filter setting will be used throughout.