src/problems/mprt.lisp @ 93ba483d9bc9
Move function
author |
Steve Losh <steve@stevelosh.com> |
date |
Sun, 25 Aug 2019 22:59:19 -0400 |
parents |
b3598dc7ef5c |
children |
2735aa6aab79 |
(in-package :rosalind)
;; This was pretty simple, except for discovering that cl-ppcre's all-matches
;; function skips overlapping matches. Otherwise we just convert the motif to
;; a regex and handle grabbing the data from Uniprot (which is straightforward
;; but can be slow).
(defparameter *input-mprt*
"A2Z669
B5ZC00
P07204_TRBM_HUMAN
P20840_SAG1_YEAST")
(defparameter *output-mprt*
"B5ZC00
85 118 142 306 395
P07204_TRBM_HUMAN
47 115 116 382 409
P20840_SAG1_YEAST
79 109 135 248 306 348 364 402 485 501 614
")
(defparameter *motif-n-glycosylation* "N{P}[ST]{P}")
(defun motif-to-regex (motif)
"Turn a protein motif shorthand into a PPCRE scanner."
(-<> motif
;; All we have to do is turn {X} into [^X] and compile.
(ppcre:regex-replace-all "[{]" <> "[^")
(substitute #\] #\} <>)
ppcre:create-scanner))
(defun all-matches-dammit (regex target-string)
"Return a list of start and end positions of all matches of `regex` on `target-string`.
Unlike `ppcre:all-matches` this will return ALL matches, even if they're
overlapping. Example:
(all-matches-dammit \"a..\" \"aabc\")
; =>
; (0 3 1 4)
(ppcre:all-matches \"a..\" \"aabc\")
; =>
; (0 3) ; dammit
"
;; cl-ppcre
(iterate
(with i = 0)
(for (values start end) = (ppcre:scan regex target-string :start i))
(while start)
(collect start)
(collect end)
(setf i (1+ start))))
(define-problem mprt (data stream)
*input-mprt*
*output-mprt*
(with-output-to-string (s)
(iterate
(with n-glycosylation = (motif-to-regex *motif-n-glycosylation*))
(for id :in-stream data :using #'read-line)
(for (nil . protein) = (uniprot id))
(for matches = (all-matches-dammit n-glycosylation protein))
(when matches
(format s "~A~%~{~D~*~^ ~}~%" id (mapcar #'1+ matches))))))