Dynamic Programming over Intervals

The Problem

高中生物知识：double-stranded DNA is “zipped” together by complementary base-pairing. Each strand of DNA can be viewed as a string of bases, where each base is drawn from the set { A , C , G , T } \{A,C,G,T\} {A,C,G,T}. The bases $A-T$ pair with each other and the bases $C-G$ pair with each other.
A single strand RNA pairs with itself, resulting in

The set of pairs (and resulting shape) formed by the RNA molecule through this process is called the secondary structure.

Single-Strand RNA Setting

Bases: { A , C , G , U } \{A,C,G,U\} {A,C,G,U}
Structure: a sequence of $n$ bases $B=b_1{b}_2,\dots ,b_n$
Rules:

• $A$ pair only with $U$, $G$ pair only with $C$.
• Bases at location $i,j$ cannot form a band if $j<i+4$ for $j>i$
• A base cannot form multiple bands
• No crossing pair is allowed. If $(i,j)$ and $(k,l)$ are two pairs, then we cannot have $i<k<j<l$

Designing and Analyzing the Algorithm

思路

尝试Greedy。从sequence里面第一个base开始寻找可以的pair。
发现此方法会形成crossing pair。
尝试Dynamics Programming 一维数组的表达形式：假设$W[i]=$max base pairs using location $1,\dots ,i$, and let $\mathcal{O}$ be an optimal pairing。那么, let $(i,n)$ be a pair in $\mathcal{O}$: if $n$ is not in the optimal pairing set $\mathcal{O}$, $W[n]=W[n-1]$ . If $n$ is in the optimal pairing, $W[n]=1+W[i-1]+$optimal pairing for strand $i+1,\dots ,,n-1$.

算法

Let $OPT(i,j)$ denote the maximum number of base pairs in a secondary structure on $b_i,\dots ,b_j$.

两种情况：

• $j$ is not involved in a pair
• $j$ pairs with $t$ for some $t<j-4$

Base Case: initialize $OPT(i,j)=0$ for $i\ge j-1$.
Inductive Case: $OPT(i,j)=\max \left(OPT(i,j-1),max(1+OPT(i,t-1)+OPT(t+1,j-1))\right)$
where $t<j-4$ and $b_t$ and $b_j$ are allowable pair (A-U OR C-G)

分析

There are $O(n^2)$ subproblems to solve, and evaluating the recurrence takes time $O(n)$ for each. Thus the running time is $O(n^3)$.

Reference

Algorithm Design (Jon Kleinberg)