2010年12月04日

DNA配列の性質

1996/7/9, posted by Robert Firth

A fascinating short article in the April 1995issue of 'Discover' (a popular science magazine)tells how a group of researches into human DNAdecided to see whether the codon sequencesfollowed Zipf's Law.

Somewhat to their surprise, they did, even in theso-called "junk" DNA (the 95%+ of human DNA thatdoesn't seem to do anything).

Their conclusion: "we don't know what it says, butit's language". Sound familiar?

1996/7/9, posted by Gabriel Landini

I 've been to a talk by of the authors of that research group (S Havlin)presenting the data in Marseille in the Fractal 95 Conference and I was not very impressed.The problem is that DNA does not have "words" and so they invented the "word" as n-base subsequences.This of course does not have anything to do with Zipf's law but to the relative probability of the bases. Yes, the DNA is different in coding and non-coding parts but this has been known for ages, so this "new finding" is not that new. The same results can be achieved with the n-base entropy, and it has a much more solid basis than "Zipf's law".

Also, you can get Zipf's distributions with absolutely random sequences.So the "language" of the junk DNA is a very far fetched hypothesis. I am not saying that it is not right, only that the evidence for having a "language" is very weak. The junk DNA is there for some reason and there are more interesting hypotheses on why we have accumulated DNA which is of no use.

The reason for the Zipf's law in random sequences is a different one from the one in the texts (well, as far as I understood W. Li's paper).Zipf's laws in texts may be important when you know that you have a "real" text and want to compare "distances" between the Zipf's plots. (I think that a reference for that is a short paper in Physical Review Letters E, by S. Havlin. If anyone is interested I can look for the reference).
posted by ぶらたん at 10:00| Comment(0) | その他
この記事へのコメント
コメントを書く
お名前:

メールアドレス:

ホームページアドレス:

コメント: [必須入力]

認証コード: [必須入力]


※画像の中の文字を半角で入力してください。
HPへ戻る