regexp with gaps

[egrasso]

Hi, I need to find the position of some substrings inside of a long
string. For this I'm using a loop that uses str.index(pattern,
(last_found_position+1)) so I find all positions where the pattern
matches. The pattern is a string of 20 chars, different each time I
run the script. That worked perfect. The problem is that now I need to
find all positions where the pattern matches 12 or more chars.
For example: For the pattern "aaaaaa", find substrings "aaaaaa",
"aaabaa", "baaaaa", "ababaa", etc

First I thought that I could create all possible patterns (with \w)
and check them, but I realized that there would be a lot of different
patterns to check (over a few hundreds I think).
Is there any way to do this without the need of checking a lot of
patterns?
thanks

 

[phlip]

The problem is that now I need to

find all positions where the pattern matches 12 or more chars.
For example: For the pattern "aaaaaa", find substrings "aaaaaa",
"aaabaa", "baaaaa", "ababaa", etc

First I thought that I could create all possible patterns (with \w)

\w{12,}

Right?

Either that or \w{12}\w*

 

[egrasso.rb]

Mmmmm... nop. I think I didn't explain the idea very well... I'm writing a
script to find specific secuences of DNA (binding sites) inside of a large
secuence of DNA (for thosse who doesn't know, DNA sequences are made of 4
diferent bases: A, T, C and G). The problem is that the binding sites don't
need to be 100% exact to work. For example, the binding site for an X
protein is "AAATTT", but the protein can also bind to the secuence "AAAGTT"
or "AACGTT" and work fine. I need to find all this sites, but the only data
I have is that "Protein X binds to AAATTT".
I finally solve the problem without using str.index nor regexp, basically,
I seek it manually:

(Note: variables are in spanish!: buscarBS=find binding site,
patron=pattern, semejanza=1 to 0, minimal similarity, cadena=string,
respuesta=answer, largo=length)

def buscarBS(patron, semejanza=0.6, cadena=@secuencia)
respuesta = ""
i = 0.0
j = 0.0
largoc = cadena.length
largop = patron.length

while i <= (largoc-largop)
j = 0.0
puntos = 0.0
subpuntos = largop * (1-semejanza)

while (j < largop) and (subpuntos > 0)
pos = i + j
if cadena[pos] == patron[j] then
puntos +=1
else
subpuntos -=1
end
j+=1
end
if (puntos / largop) >= semejanza then
respuesta = respuesta + "desde: "+(i+1).to_i.to_s+" hasta:
"+(i+j).to_i.to_s+" - similitud: - "+(puntos / largop * 100).to_s+"%\n"
end
i+=1
end

if respuesta == "" then
respuesta = "No se encontro ninguna secuencia similar (similitud:
#{semejanza} - #{patron})"
else
respuesta = "\nSe encontraron las siguientes similitudes:\n\n"+respuesta
end
return respuesta

end

I still need to polish and optimize the code but it find all possible
sites with at least an specific similarity and tells me how similar they
are. If anyone have another idea, need more details about the code or is
interested in bioinformatic with ruby tell me.
Thanks

 

[Axel Etzold]

-------- Original-Nachricht --------

Datum: Tue, 15 Jul 2008 12:18:09 +0900
Von: (e-mail address removed)
An: (e-mail address removed)
Betreff: Re: regexp with gaps

Mmmmm... nop. I think I didn't explain the idea very well... I'm writing a
script to find specific secuences of DNA (binding sites) inside of a large
secuence of DNA (for thosse who doesn't know, DNA sequences are made of 4
diferent bases: A, T, C and G). The problem is that the binding sites
don't
need to be 100% exact to work. For example, the binding site for an X
protein is "AAATTT", but the protein can also bind to the secuence
"AAAGTT"
or "AACGTT" and work fine. I need to find all this sites, but the only
data
I have is that "Protein X binds to AAATTT".
I finally solve the problem without using str.index nor regexp, basically,
I seek it manually:

(Note: variables are in spanish!: buscarBS=find binding site,
patron=pattern, semejanza=1 to 0, minimal similarity, cadena=string,
respuesta=answer, largo=length)

def buscarBS(patron, semejanza=0.6, cadena=@secuencia)
respuesta = ""
i = 0.0
j = 0.0
largoc = cadena.length
largop = patron.length

while i <= (largoc-largop)
j = 0.0
puntos = 0.0
subpuntos = largop * (1-semejanza)

while (j < largop) and (subpuntos > 0)
pos = i + j
if cadena[pos] == patron[j] then
puntos +=1
else
subpuntos -=1
end
j+=1
end
if (puntos / largop) >= semejanza then
respuesta = respuesta + "desde: "+(i+1).to_i.to_s+" hasta:
"+(i+j).to_i.to_s+" - similitud: - "+(puntos / largop * 100).to_s+"%\n"
end
i+=1
end

if respuesta == "" then
respuesta = "No se encontro ninguna secuencia similar (similitud:
#{semejanza} - #{patron})"
else
respuesta = "\nSe encontraron las siguientes
similitudes:\n\n"+respuesta
end
return respuesta

end

I still need to polish and optimize the code but it find all possible
sites with at least an specific similarity and tells me how similar they
are. If anyone have another idea, need more details about the code or is
interested in bioinformatic with ruby tell me.
Thanks

\w{12,}

Right?

Either that or \w{12}\w*

Hi ---

you could make use of the McIlroy-Hunt longest common subsequence (LCS) algorithm,
which will give you longest common subsequences, and also information of the type

'sequence AAATTT is transformed into AAAGTT by changing T to G at the fourth entry.'

You can find a Ruby gem implementation here: http://raa.ruby-lang.org/project/diff-lcs/

Best regards,

Axel