Understanding Systems Biology: Minds among the Molecules

學期中的每個禮拜五中午都有 SysBio Dept. Theory Lunch，請一些不是做傳統 bench work 的人來講述他們的研究方向與成果。通常我只要有空就會去，一方面是瞭解一下其他人在做甚麼；另一方面提供免費午餐何樂而不為？也因此通常我不會特別查看這禮拜是請誰來講，只管拿了吃的坐著聽便是了。

今天一進門就看到 Marc Kirschner 和 Chris Walsh 這兩個加起來大概超過一百三十歲的老頭坐在裡面聊天，Marc 雖然少來，畢竟是 department chair，看到他還可以理解，Chris 搞化學與生化的來聽就很奇怪，想是演講者大概來頭不小的樣子。後來我才發現原來今天講的人是另一個我也常碰見的老頭，這一年來時常看到他在 department 內走動，始終不清楚他的來歷只知道他正在寫一本書，結果原來他就是 Carl Pabo。

Carl 今天講的就是他這幾年來的研究內容，大致上就是思考 Systems Biology 這個新的領域將來應該有的發展方向與架構。簡單說就是由於現在的 biology data >>>>> long-term memory >>>>> short-term memory，研究任何一個生物機制所需要的知識與訊息量都遠超過一個人所能記得的極限，是故生物界，或說系統生物界，將來勢必不能依靠個別實驗室的單打獨鬥來解決生物問題，而如何合作才能產生最大的效益便是他所思考的問題。他認為該模仿大腦的運作模式，而神經元如何溝通以及傳送訊息的格式則扮演了關鍵的角色，之後拉里拉雜地講了許多細節。整體來說他的演講技巧是很不錯的，今天帶了一整個推車的書供他在引述，大概有二十多本。不過或許內容實在是太玄了些，我坐在 Marc 和 Chris 的右後方，大概演講開始二十分鐘後兩個人便顯示出沒有繼續 follow 下去的徵狀，Chris 一直輪流將手上的 Diet Coke / Starbucks coffee 拿在手上再放回地上，Marc 則根本就睡著了。

語言是 Carl 提及很重要的一個例子，人們在聽到一串由單字所組成的句子時，如何由不同神經元把聽到的訊息單字重組解譯成有意義的句子，席間他並引用了 Norm Chomsky 的著作。我有點意外他竟然會引用Chomsky 來解釋他的想法，Chomsky 是 MIT 一位很傑出的語言學家，不過後期在 Chromsky 跨入認知心理的領域開始提出一連串激進的主張如天賦知識論後，幾乎與所有人如皮亞傑 (Piaget) 派、行為主義學派、訊息處理等學派都發生衝突，被狗幹得很慘。雖然 Chomsky 的某些觀點如大腦先天的構造與發展過程影響了認知能力的進程現在已經證實，但許多激進的主張就算以現今的眼光來看還是很誇張的。回到正題，Carl 主要的意思是說人類的大腦對訊息的處理有一定的格式，也就是現今所有已發展出的語言所依賴的文法，如果將這些格式加以改變或在語句的中間部分加上大量贅字或贅辭，同樣一個句子被理解的程度便會大量降低。

他接著補了一句，這就是為甚麼詩會讓人那麼難懂的原因。

我想的是，這也許就是詩意這種奇妙感覺的由來吧。

下面是 Carl Pabo Theory Lunch 的 Abstract:

Understanding systems biology: minds among the molecules
20 October 2006

Carl Pabo
Department of Systems Biology, HMS
National Academy of Sciences Bio
Stanford Bio-X Bio

Abstract
This seminar builds on ideas that I developed during my recent Guggenheim Fellowship for work on "Theories of Thought", and considers how some of these issues may be relevant in planning overall directions for the field of systems biology. My argument will proceed in three stages:

I'll start by considering several basic aspects of the function of the mind/brain (including issues such as limits of short-term memory, ambiguity and context-dependence in the meaning of symbols, use and limits of strategies for abstraction and "chunking", etc.).
I'll then briefly introduce a range of models that have been used in systems biology (including dynamical systems simulations, network motifs, theories of noise, scale-free networks, etc.) and will point out some significant differences in the cognitive, computational, and informational correlates of the various ways that biological information is represented in these different models.
I'll close by considering how the differing cognitive/informational characteristics of the models may help to determine which types of models will be most useful to the broader biomedical community [which is, itself, comprised of scientists working under the same types of cognitive constraints mentioned at the start of the talk].
Obviously, the scope of the issues raised here means that one cannot be entirely precise ("fully scientific") in representing each aspect of this argument. However, the analysis builds on a rich foundation of data from the cognitive neurosciences, and several constraints seem so clear and so striking that I expect that they will have profound practical implications for the optimal development of this new field.