STATISTICAL MECHANICS OF PROTEIN FOLDING

关于蛋白质折叠的统计力学已经提出了许多理论，这些理论通常来自于结构较少的系统的理论，如随机异聚物、扩散成核、随机能量或自旋玻璃。虽然这些方法确实捕捉到了真实蛋白质的一些重要物理化学特征，比如折叠的协同性和随机线圈状态下的快速折叠，但它们的推导需要对多肽的平均行为做出一些广泛的假设。事实上，具有生物学意义的蛋白质显然是由非常罕见的、中等长度的氨基酸序列组成的，这些氨基酸序列允许链迅速折叠，形成独特复杂的天然构象，这在很大程度上取决于序列。这表明，关注长链杂聚物的平均性质的理论可能过于一般化，而忽视了罕见序列折叠成罕见构象的重要特征。另一方面，自然界和计算机都没有足够的时间来详尽地探索即使是小蛋白质的所有构象和所有序列。这里的想法是通过减少链长，每个残基的构象状态数量和氨基酸类型的选择，直到所有序列和所有构象都可以被详尽地列举出来，从而简单地表示多肽。通过在计算可行范围内改变这些参数，可以检测到一般结论并推断出与实际蛋白质对应的参数值。由于这个模型与大多数理论如此不同，它能够测试他们关于蛋白质折叠的假设和结论，例如能量景观的性质和描述向原生状态进展的顺序参数。要解决的问题包括：是否有一种通用的方法来描述所有蛋白质的折叠，或者一些蛋白质通过可识别的途径进行，而另一些蛋白质有无数的途径？这个模型能否再现和解释目前关于某些特定蛋白质的折叠机制和中间体的实验结果？

Many theories have been proposed for the statistical mechanics of protein folding, typically derived from theories for less structured systems, such as random heteropolymers, diffusion-nucleation, random energies, or spin glasses. While these have certainly captured some important features of the physical chemistry of real proteins, such as cooperativity of folding and rapid folding from the random coil state, their derivation requires making some broad assumptions about the average behavior of polypeptides. In fact, the proteins of biological relevance consists of apparently very rare, moderately long amino acid sequences that permit the chain to fold rapidly- to a unique complicated native conformation, which depends greatly on the sequence. This suggests that theories focussing on average properties of long-chain heteropolymers may be over-generalizing and neglecting the important features of rare sequences folding to rare conformations. On the other hand, neither nature nor computer has sufficient time to exhaustively explore all conformations and all sequences for even small proteins. The idea here is to simply an otherwise realistic representation of polypeptides by reducing chain length, number of conformation states per residue, and choices of amino acid types until all sequences and all conformations can be exhaustively enumerated. By varying these parameters in the computationally feasible range, general conclusions can be detected and extrapolated to parameter values corresponding to real proteins. Since this model is so different from most theories, it is able to test their assumptions and conclusions about protein folding, such as the nature of the energy landscape and order parameters to describe the progress toward the native state. Questions to be addressed include: is there a general way to describe the folding of all proteins, or do some proceed by a recognizable pathway while others have innumerable routes? Can this model reproduce and explain the currently available experimental results on folding mechanisms and intermediates for certain particular proteins.