English

Life processes information. A cell reads the instructions encoded in its DNA, transcribes and translates them into proteins, regulates which instructions are active in response to its environment, and passes a copy of the instructions to its descendants. This is information processing, performed by molecular machinery, and the recognition that life is at bottom an informational phenomenon is the deepest root of the relationship between computer science and biology. DNA is a digital code — a sequence over a four-letter alphabet — and the machinery that reads and acts on it is, in a meaningful sense, a computational system. The discovery of this, in the molecular biology revolution of the mid-twentieth century, established that the concepts of information and computation are not merely tools for studying life from the outside but are descriptive of what life is.

The relationship has three faces, each substantial. Life as information processing is the deepest: the recognition that biological systems are information-processing systems, that the central dogma of molecular biology (DNA to RNA to protein) is a description of information flow, and that the concepts of computation illuminate what living systems are doing. Computational biology is the most practically consequential: the application of computer science to biological problems, which has become essential to biology — the sequencing and analysis of genomes, the prediction of protein structure, the modeling of biological systems are now computational activities, and biology has become, in significant part, a computational science. And biologically inspired computation is the reverse flow: biology has inspired computational methods — evolutionary algorithms drawn from natural selection, neural networks drawn from the brain, and more speculative directions like DNA computing — that bring biological principles into computer science.

This section examines all three: life as an information-processing phenomenon, computational biology as the application of computation to life, and biologically inspired computation as the flow of biological ideas into computing. It connects to machine learning (§5.2–§5.3, increasingly central to computational biology), to algorithms (§2.6, the basis of sequence analysis), and to information theory (§3.8), and it engages biology that is outside this guide’s scope but whose relationship to computation is the subject.

Background: Algorithms (§2.6), machine learning (§5.2–§5.3), and information theory (§3.8) provide the computational frame; molecular biology, genetics, evolution, and systems biology provide the biological one.

The Relationship: Life, Computation, and Their Mutual Illumination

Life as Information Processing

The deepest insight of the relationship is that life is, fundamentally, an information-processing phenomenon, and that the discovery of this was the molecular biology revolution. The structure of DNA, determined by Watson and Crick in 1953, revealed that genetic information is stored in a digital code — a linear sequence over a four-letter alphabet (the bases A, T, G, C) — and the subsequent decade established how this code is read and acted upon. The central dogma of molecular biology (Crick, 1958) — that information flows from DNA to RNA to protein — is, explicitly, a description of information flow, and the language of the field (code, transcription, translation, reading frame, expression) is the language of information processing, not by loose analogy but because that is what the processes are.

The genetic code itself is a code in the precise sense: a mapping from sequences of nucleotides (read in triplets, codons) to amino acids, with the properties of a code — redundancy (multiple codons map to the same amino acid), error characteristics, start and stop signals. That life uses a digital code, rather than analog chemistry, is itself significant: digital encoding allows information to be copied with high fidelity and to be stored stably, which are prerequisites for heredity, and the digital nature of the genetic code is part of why life can maintain and transmit information across generations. The recognition that heredity is the transmission of digital information was one of the great unifications of twentieth-century science, bringing the concept of information to the center of biology.

The information-processing view extends beyond the genetic code to the whole of cellular function. Gene regulation — the control of which genes are active when — is information processing: the cell senses its state and environment and computes, through networks of interacting molecules, which genes to express, in a way that has been productively modeled as computation (gene regulatory networks as circuits, signaling pathways as information channels). Cells make decisions, process signals, and respond to their environment through molecular networks that implement what are recognizably computational functions. Systems biology, the study of these networks, uses the concepts and tools of computer science (network analysis, dynamical systems, information theory) to understand the information processing that cells perform. The view of the cell as an information-processing system — sensing, computing, deciding, acting — is a productive framework that the relationship between the fields made possible, and it represents the deepest sense in which computation illuminates life.

Computational Biology: Computation in the Service of Life Science

Computational biology — the application of computer science to biological problems — has become essential to biology, transforming it into a science that cannot be done without computation. The transformation is recent and dramatic: within a few decades, biology has gone from a science where computation was peripheral to one where the central activities of generating and analyzing data are computational.

The transformation was driven by sequencing. The ability to read DNA sequences, and the explosive growth in the speed and cheapness of sequencing (far outpacing Moore’s law), produced biological data at a scale that only computation could handle. The Human Genome Project (completed 2003) was a computational as much as a biological achievement, requiring algorithms to assemble the genome from fragments and to make sense of the result. Bioinformatics — the development of algorithms and tools for biological data — became a field: sequence alignment (the BLAST algorithm and its successors, finding similarities between sequences), genome assembly, the identification of genes and regulatory elements, the comparison of genomes across species. These are algorithmic problems (string matching, dynamic programming, graph algorithms applied to biological sequences), and the algorithms of §2.6 are the foundation of the field. Biology became, in significant part, a data science, and the computational tools became as essential to it as the laboratory tools.

The most striking recent achievement is the prediction of protein structure, where machine learning has dramatically advanced a problem that resisted biology for fifty years. Predicting the three-dimensional structure a protein adopts from its amino acid sequence was a grand challenge of molecular biology, essential because structure determines function and experimental determination of structure is slow and expensive. DeepMind’s AlphaFold (2020, §6.5) predicted many protein structures with accuracy comparable to experiment, using a deep learning model trained on known structures. This was a transformation: a central part of a fifty-year-old problem became practically tractable through machine learning, with enormous consequences for biology and medicine (understanding disease, designing drugs, engineering proteins). It exemplifies the deepest mode of computational biology — not just analyzing biological data but solving fundamental biological problems with computational methods — and it points toward a future in which machine learning is central to biological discovery. The broader application of machine learning to biology (predicting the effects of genetic variants, designing proteins and genetic sequences, modeling cellular systems, drug discovery) is among the most active and consequential frontiers in both fields.

Biologically Inspired Computation: Life’s Ideas in Computing

The reverse flow — biological principles inspiring computational methods — is the third face of the relationship, and it runs from well-established methods to speculative frontiers. Biology has been a source of inspiration for computer science because evolution has produced solutions to problems (optimization, learning, robustness, self-organization) that computer science wants to solve, and abstracting biological principles into computational methods has been productive.

Evolutionary computation abstracts the principle of natural selection into an optimization method: maintain a population of candidate solutions, evaluate their fitness, select the fitter ones to reproduce (with variation through mutation and recombination), and iterate, so that the population evolves toward better solutions. Genetic algorithms, genetic programming, and evolution strategies apply this principle to optimization problems where the structure is poorly understood and gradient-based methods do not apply. Evolutionary computation is a real tool — useful for certain optimization and design problems, particularly where the search space is complex and the objective is not differentiable — and it is a direct importation of a biological principle (selection acting on variation) into computation.

Neural networks are the most consequential biologically inspired computation, though, as §9.3 discussed, the inspiration is loose. The principle abstracted from biology — that intelligence can arise from networks of simple units adjusting their connections — was the seed, and the brain was the existence proof, but the artificial neural networks that became deep learning diverged far from biological neurons. Still, the biological inspiration was the origin, and the success of deep learning is, at the root, the vindication of a biologically inspired idea, however far it has traveled from its biological source.

The more speculative frontiers explore deeper fusions. DNA computing (Adleman’s 1994 demonstration that a computational problem could be solved using DNA molecules) explores using biological molecules as a computational substrate, exploiting the massive parallelism of molecular reactions; it is more a proof of concept and a research curiosity than a practical technology, but it demonstrates that computation is substrate-independent in a striking way. Synthetic biology — the engineering of biological systems, designing genetic circuits that implement desired functions — applies engineering and computational principles to biology, treating cells as programmable systems and DNA as code to be written, not just read. The vision of programming cells, of designing biological systems as one designs software, is a frontier where the information-processing view of life meets the engineering ambition of computer science, and it represents the relationship at its most forward-looking — the ambition not just to read life’s code but to write it.

What This Perspective Changes

Seeing the relationship between computer science and biology changes how a practitioner understands both computation and life.

The first change is the recognition that life is informational. The practitioner who grasps that DNA is a digital code, that the central dogma describes information flow, that cells perform information processing, understands one of the deepest facts about life — that it is, at bottom, an informational phenomenon — and sees that the concepts of computer science are not merely tools applied to biology from outside but are descriptive of what life is. This is a profound reframing of what living systems are, available through the relationship between the fields.

The second change is appreciation of how thoroughly computation has transformed biology. The practitioner who sees that biology has become, in significant part, a computational science — that sequencing produces data only computation can handle, that protein structure is now predicted by machine learning, that biological discovery increasingly proceeds through computation — understands a major transformation in a major science, and one in which computer science skills are increasingly central to biological research. For a computationally trained person, biology has become a field where their skills are not only applicable but essential, and one of the most consequential places to apply them.

The third change is calibrated understanding of biologically inspired computation. The practitioner who understands the flow of biological ideas into computing — evolutionary computation, neural networks, the speculative frontiers — can assess these methods correctly: evolutionary computation as a real but specialized tool, neural networks as a loosely biologically inspired idea that succeeded by diverging from biology, the speculative directions as frontiers of varying maturity. This calibration avoids both the overclaiming (biological inspiration as a guarantee of success) and the dismissal (biological inspiration as mere metaphor).

The fourth change is a glimpse of the most consequential frontier. The practitioner who sees the convergence — machine learning solving fundamental biological problems, synthetic biology treating cells as programmable, the information-processing view of life meeting the engineering ambition of computer science — glimpses what may be among the most consequential developments of the coming decades. The ability to read, understand, and increasingly write the information that constitutes life, using the tools of computation, is a frontier with implications for medicine, for our understanding of life, and for what becomes possible, and the relationship between the fields is where it is happening.

Resources

Foundational and Bridging Texts

For the information-processing view of life, the foundational presentation is in the molecular biology that established it; The Molecular Biology of the Gene (Watson and others, now in its 7th edition) and Molecular Biology of the Cell (Alberts and others) are the standard texts where the informational nature of life is presented, though as biology texts rather than as treatments of the relationship. For the relationship itself — life as information — the writing that frames biology informationally is the entry: Schrödinger’s What Is Life? (1944), remarkably, anticipated the idea of a genetic code before its discovery and influenced the founders of molecular biology, and remains a profound short reflection on the relationship between physics, information, and life.

For computational biology as a field, Compeau and Pevzner’s Bioinformatics Algorithms: An Active Learning Approach (and the associated Rosalind platform, rosalind.info, free) is the best entry for a computationally trained person: it teaches the algorithms of bioinformatics (sequence alignment, assembly, motif finding) through problems, connecting the computer science of §2.6 to the biological applications. Durbin, Eddy, Krogh, and Mitchison’s Biological Sequence Analysis (1998) is the classic rigorous treatment of the probabilistic models (hidden Markov models, and more) central to sequence analysis.

For the broad relationship, accessible books on the convergence — including treatments of how machine learning is transforming biology and medicine — provide the contemporary picture, though the field moves fast enough that the primary literature is where the frontier is.

Resource Role Type
Schrödinger, What Is Life? (1944) Life as information; prescient foundation Depth
Compeau & Pevzner, Bioinformatics Algorithms + Rosalind (free) Computational biology entry for CS people Practice
Durbin et al., Biological Sequence Analysis Probabilistic models for sequences Depth
Alberts et al., Molecular Biology of the Cell The biology, where information framing lives Reference

Computational Biology, AI, and Biologically Inspired Methods

For the machine learning transformation of biology, the AlphaFold papers (Jumper et al., 2021, free) are the primary source for the landmark achievement, and reading them alongside the §6.5 and §5.3 material connects the deep learning methods to the biological problem. The broader literature on machine learning in biology and medicine is developing rapidly; following it requires the primary literature and the reviews, as the field is moving faster than books.

For biologically inspired computation, Mitchell’s An Introduction to Genetic Algorithms (1996) is the accessible introduction to evolutionary computation, and the broader literature on bio-inspired computing covers the range of methods. For DNA computing and synthetic biology, Adleman’s original paper (1994, free) is the readable origin of DNA computing, and the synthetic biology literature (including the writing on engineering biology as programming) covers the frontier of writing biological code.

For the deepest treatments of the convergence — the information-theoretic understanding of biology, the computational view of evolution and life — the work of researchers bridging the fields (the systems biology literature, the work on the thermodynamics and information theory of living systems) is the frontier, largely in the primary literature.

Resource Role Type
Jumper et al., AlphaFold papers (2021, free) The protein-structure prediction landmark Depth
Mitchell, An Introduction to Genetic Algorithms Evolutionary computation entry Entry
Adleman, DNA computing paper (1994, free) DNA computing origin Depth
ML-in-biology literature (current) The active frontier Reference
AlphaFold Protein Structure Database (free) Open protein-structure predictions at scale Practice
EMBL-EBI AlphaFold practical guide (free) Practical guide to using AlphaFold and reading confidence Practice
Rosalind platform (free) Bioinformatics through problems Practice

Traps

Trap Why it misleads Better response
Treating “DNA is code” as mere metaphor The description of DNA as code and the cell as a computer can sound like loose analogy, a way of talking about biology in borrowed terms. It is more than that: DNA genuinely is a digital code in the precise sense (a sequence over an alphabet, read and translated by molecular machinery), and the information-processing view of life is descriptive, not metaphorical. Dismissing it as metaphor misses one of the deepest facts about life. Take the information-processing view of life literally where it is literal. DNA is a digital code; the central dogma describes information flow; cells process information through molecular networks. The concepts of computation are descriptive of what living systems do, not borrowed metaphors. At the same time, hold the appropriate caution about where the analogy stretches (the cell is not a von Neumann computer), distinguishing the literal informational nature of life from looser computational analogies.
Overstating the biology-to-CS inspiration Biologically inspired methods (evolutionary computation, neural networks) are sometimes promoted as superior because they are biologically inspired, as though drawing on biology guaranteed success. The inspiration is a source of ideas, not a guarantee of quality, and the most successful biologically inspired method (neural networks) succeeded by diverging far from its biological source. Biological inspiration is a heuristic for generating methods, not a validation of them. Evaluate biologically inspired methods on their merits, not on their biological pedigree. Evolutionary computation is useful for specific problems (complex, non-differentiable search spaces) and not a general-purpose superior method; neural networks succeeded by abandoning biological fidelity for what worked. The biological inspiration is where the ideas came from; whether they work is a separate question answered by their performance, not their origin.
Underestimating how computational biology has become A person trained in computer science may regard biology as a distant field with little connection to their skills. This is increasingly wrong: biology has become, in significant part, a computational science, and computational skills are central to modern biological research — in genomics, in structural biology, in systems biology, in drug discovery. The computational person who dismisses biology as unrelated to their skills misses one of the largest and most consequential domains where those skills apply. Recognize biology as a domain where computational skills are increasingly essential and the problems are among the most consequential. The transformation of biology into a data-and-computation-intensive science means that computationally trained people have a major and growing role in biological research. For someone deciding where to apply computational skills, the convergence with biology is among the most important and impactful frontiers.

中文

生命处理信息。一个细胞读取编码在 DNA 中的指令，将其转录并翻译成蛋白质，根据环境调节哪些指令处于活跃状态，并把这些指令的副本传递给后代。这就是信息处理，由分子机器完成。认识到生命在根本上是一种信息现象，是计算机科学与生物学关系最深的根源。DNA 是一种数字代码——由四个字母组成的序列——而读取并执行它的机制，在有意义的层面上，就是一个计算系统。20 世纪中叶分子生物学革命发现了这一点，并确立了这样一个事实：信息和计算的概念并不只是从外部研究生命的工具，它们描述的正是生命本身。

这种关系有三种面向，每一种都很重要。第一种也是最深的一种，是把生命理解为信息处理：生物系统是信息处理系统，分子生物学的中心法则——DNA 到 RNA 到蛋白质——描述的是信息流，而计算概念能够照亮生命系统正在做什么。第二种是计算生物学，它在实践上最具后果：把计算机科学应用于生物学问题，并且已经成为生物学不可或缺的一部分。基因组测序与分析、蛋白质结构预测、生物系统建模，如今都是计算活动；生物学在很大程度上已经成为一门计算科学。第三种是反向流动，也就是受生物启发的计算：生物学启发了计算方法，例如来自自然选择的进化算法、来自大脑的神经网络，以及 DNA 计算这类更具探索性的方向，它们把生物原则带入计算机科学。

本节将考察这三者：作为信息处理现象的生命、作为计算应用于生命科学的计算生物学，以及作为生物思想流入计算的受生物启发的计算。它连接到机器学习（§5.2–§5.3，正在计算生物学中变得越来越核心）、算法（§2.6，序列分析的基础），以及信息论（§3.8）；同时也涉及本指南范围之外的生物学，而它与计算的关系正是本节主题。

背景知识：算法（§2.6）、机器学习（§5.2–§5.3）和信息论（§3.8）提供计算框架；分子生物学、遗传学、进化论和系统生物学提供生物学框架。

这种关系：生命、计算及其相互照亮

生命作为信息处理

这种关系最深的洞见是：生命在根本上是一种信息处理现象；而这一点的发现，正是分子生物学革命的核心。Watson 和 Crick 于 1953 年确定 DNA 结构，揭示了遗传信息储存在一种数字代码中——由四种碱基 A、T、G、C 组成的线性序列。随后的十年则说明了这套代码如何被读取并发挥作用。分子生物学的中心法则（Crick，1958）——信息从 DNA 流向 RNA，再流向蛋白质——明确就是对信息流的描述；而这个领域使用的语言，例如代码、转录、翻译、阅读框、表达，也都是信息处理的语言。这不是松散类比，而是因为这些过程本身就是如此。

遗传密码本身就是严格意义上的代码：它把核苷酸序列，也就是按三联体读取的密码子，映射到氨基酸上，并具有代码的性质——冗余性，即多个密码子可以对应同一个氨基酸；错误特征；起始和终止信号。生命使用的是数字代码，而不是模拟化学过程，这一点本身就很重要：数字编码使信息能够以高保真度复制，并被稳定储存，而这些正是遗传的前提。遗传密码的数字性质，也是生命能够跨世代维持并传递信息的原因之一。认识到遗传就是数字信息的传递，是 20 世纪科学最重要的统一之一，它把信息概念带到了生物学的中心。

信息处理视角并不止于遗传密码，而是延伸到整个细胞功能。基因调控——控制哪些基因在什么时候活跃——就是信息处理：细胞感知自身状态和环境，并通过相互作用的分子网络进行计算，决定表达哪些基因。这个过程已经被富有成效地建模为计算，例如把基因调控网络看作电路，把信号通路看作信息通道。细胞通过分子网络做出决策、处理信号并响应环境，而这些分子网络实现的是可以明确识别为计算功能的活动。系统生物学研究这些网络，并使用计算机科学的概念和工具——网络分析、动力系统、信息论——来理解细胞所执行的信息处理。把细胞看作信息处理系统——感知、计算、决策、行动——是两个领域关系所开启的一个富有成效的框架，也代表了计算照亮生命的最深含义。

计算生物学：服务生命科学的计算

计算生物学——把计算机科学应用于生物学问题——已经成为生物学不可或缺的一部分，并把生物学转变成一门离不开计算的科学。这种转变既新近又剧烈：在短短几十年间，生物学已经从一门计算处于边缘位置的科学，变成一门其核心活动——生成数据和分析数据——都依赖计算的科学。

推动这种转变的是测序。读取 DNA 序列的能力，以及测序速度和成本的爆炸式改善——其速度远超摩尔定律——产生了只有计算才能处理的生物数据规模。人类基因组计划（2003 年完成）既是生物学成就，也是计算成就：它需要算法把片段组装成基因组，并理解得到的结果。生物信息学，即为生物数据开发算法和工具的领域，由此形成：序列比对，例如 BLAST 算法及其后继者，用来寻找序列之间的相似性；基因组组装；基因和调控元件识别；跨物种基因组比较。这些都是算法问题：字符串匹配、动态规划、应用于生物序列的图算法；而 §2.6 中的算法正是这个领域的基础。生物学在很大程度上变成了一门数据科学，计算工具也变得和实验室工具一样不可或缺。

最近最惊人的成就是蛋白质结构预测。机器学习极大推进了一个困扰生物学五十年的问题：从蛋白质的氨基酸序列预测其采用的三维结构。这是分子生物学中的重大挑战，因为结构决定功能，而通过实验确定结构既缓慢又昂贵。DeepMind 的 AlphaFold（2020，§6.5）使用在已知结构上训练的深度学习模型，预测了许多蛋白质结构，其准确度可与实验结果相比。这是一场转变：一个持续五十年的核心问题，有相当一部分因为机器学习而变得实践上可处理，并对生物学和医学产生巨大影响，例如理解疾病、设计药物、工程化蛋白质。它体现了计算生物学最深层的模式：不只是分析生物数据，而是用计算方法解决基础生物学问题；同时也指向一个未来，在这个未来中，机器学习将成为生物发现的核心。机器学习在生物学中的更广泛应用——预测遗传变异影响、设计蛋白质和遗传序列、建模细胞系统、药物发现——已经成为两个领域中最活跃也最具后果的前沿之一。

受生物启发的计算：生命思想进入计算

反向流动——生物原则启发计算方法——是这种关系的第三种面向。它从已经成熟的方法一直延伸到更具猜想性的前沿。生物学之所以成为计算机科学的灵感来源，是因为进化已经为许多计算机科学也想解决的问题提供了解法：优化、学习、鲁棒性、自组织。把生物原则抽象成计算方法，已经产生了富有成效的成果。

进化计算把自然选择原则抽象成一种优化方法：维持一组候选解，评估它们的适应度，选择更优者进行繁殖，并通过突变和重组引入变化，然后反复迭代，使整个群体向更好的解演化。遗传算法、遗传编程和进化策略，都把这一原则应用于那些结构理解不充分、梯度方法无法适用的优化问题。进化计算是一种真实工具——对于某些优化和设计问题有用，尤其是在搜索空间复杂且目标不可微的情况下——它也是把生物原则，也就是选择作用于变异，直接引入计算的例子。

神经网络是最具后果的受生物启发的计算方法，尽管如 §9.3 所讨论的那样，它的生物启发相当松散。从生物学中抽象出来的原则是：智能可以从许多简单单元组成的网络中产生，而这些单元会调整彼此之间的连接。大脑提供了存在性证明，而这个原则成为种子；但最终发展成深度学习的人工神经网络，已经远远偏离了生物神经元。尽管如此，生物启发仍是其源头；深度学习的成功，在根本上也证明了一种受生物启发的思想可以成功，尽管它已经离自己的生物源头走得很远。

更具探索性的前沿，则在研究更深层的融合。DNA 计算——Adleman 于 1994 年展示可以用 DNA 分子解决一个计算问题——探索把生物分子作为计算基底，利用分子反应的大规模并行性。它更多是一种概念验证和研究奇观，而非实用技术，但它以一种醒目的方式展示了计算的基底无关性。合成生物学——对生物系统进行工程化，设计能实现特定功能的遗传电路——则把工程和计算原则应用于生物学，把细胞视为可编程系统，把 DNA 视为可以书写、而不仅仅是读取的代码。编程细胞、像设计软件一样设计生物系统，这一愿景正是生命的信息处理视角与计算机科学的工程雄心相遇之处；它代表了这种关系最面向未来的一端——不只是读取生命代码，而是书写生命代码。

这种视角会改变什么

看清计算机科学与生物学的关系，会改变实践者对计算和生命的理解。

第一种改变，是认识到生命具有信息性。实践者如果理解 DNA 是数字代码，中心法则描述的是信息流，细胞执行信息处理，就会理解关于生命最深的事实之一：生命在根本上是一种信息现象。由此也会看见，计算机科学概念并不只是从外部应用于生物学的工具，而是在描述生命本身。这是对生命系统是什么的一次深刻重构，而这种重构正是通过两个领域的关系获得的。

第二种改变，是理解计算已经多么彻底地改变了生物学。实践者如果看见生物学在很大程度上已经成为一门计算科学——测序产生的数据只能由计算处理，蛋白质结构如今可以由机器学习预测，生物发现越来越多地通过计算推进——就会理解一门主要科学正在发生的重大转变，而且计算机科学技能正日益成为生物学研究的核心。对于受过计算训练的人来说，生物学已经成为一个他们的技能不仅适用、而且不可或缺的领域，也成为应用这些技能最具后果的地方之一。

第三种改变，是对受生物启发的计算获得校准过的理解。实践者如果理解生物思想如何流入计算——进化计算、神经网络以及更具探索性的前沿——就能正确评估这些方法：进化计算是真实但专门化的工具；神经网络是一个受生物松散启发、并通过偏离生物而成功的思想；那些更具探索性的方向，则是成熟度各不相同的前沿。这种校准能够避免两种错误：一种是过度主张，把生物启发当成成功保证；另一种是直接否定，把生物启发看作只是隐喻。

第四种改变，是看见最具后果的前沿。实践者如果看见这种汇合——机器学习正在解决基础生物学问题，合成生物学把细胞视为可编程对象，生命的信息处理视角与计算机科学的工程雄心正在相遇——就能瞥见未来几十年中可能最具后果的发展之一。借助计算工具读取、理解，并越来越多地书写构成生命的信息，这一前沿将影响医学、我们对生命的理解，以及未来什么会变得可能；而计算机科学与生物学的关系，正是这一切发生的地方。

资源

基础与桥接文本

关于生命的信息处理视角，其基础呈现在建立这一视角的分子生物学中；The Molecular Biology of the Gene（Watson 等，现为第 7 版）和 Molecular Biology of the Cell（Alberts 等）是标准教材，其中展示了生命的信息性质，尽管它们是生物学教材，而不是专门处理计算机科学与生物学关系的文本。至于这种关系本身——生命作为信息——以信息视角讨论生物学的写作是入口：Schrödinger 的 What Is Life?（1944）在遗传密码被发现之前就惊人地预见了这一思想，并影响了分子生物学的奠基者；它至今仍是关于物理学、信息与生命关系的一篇深刻短论。

对于作为领域的计算生物学，Compeau 和 Pevzner 的 Bioinformatics Algorithms: An Active Learning Approach，以及配套的 Rosalind 平台（rosalind.info，免费），是受过计算训练者的最佳入口。它通过问题讲授生物信息学算法，例如序列比对、组装、motif 发现，把 §2.6 中的计算机科学算法连接到生物学应用。Durbin、Eddy、Krogh 和 Mitchison 的 Biological Sequence Analysis（1998）是关于序列分析中核心概率模型，例如隐马尔可夫模型等的经典严谨处理。

关于这种关系的整体图景，讨论计算机科学与生物学汇合的通俗书籍——包括机器学习如何改变生物学和医学的著作——提供了当代图景；不过该领域发展速度足够快，真正的前沿仍然在一手文献中。

资源 作用 类型
Schrödinger，What Is Life?（1944） 生命作为信息；具有预见性的基础文本 深入
Compeau & Pevzner，Bioinformatics Algorithms + Rosalind（免费） 面向 CS 学习者的计算生物学入口 实践
Durbin 等，Biological Sequence Analysis 序列概率模型 深入
Alberts 等，Molecular Biology of the Cell 生物学本身，信息框架所在之处 参考

计算生物学、AI 与受生物启发的方法

关于机器学习对生物学的改变，AlphaFold 论文（Jumper 等，2021，免费）是这一里程碑成就的一手来源。将其与 §6.5 和 §5.3 的材料一起阅读，可以把深度学习方法与生物学问题连接起来。关于机器学习在生物学和医学中的更广泛文献正在迅速发展；由于这个领域发展快于书籍，跟进它需要阅读一手文献和综述。

关于受生物启发的计算，Mitchell 的 An Introduction to Genetic Algorithms（1996）是进化计算的易读入门；更广泛的 bio-inspired computing 文献则覆盖相关方法范围。关于 DNA 计算和合成生物学，Adleman 的原始论文（1994，免费）是 DNA 计算的可读起点；合成生物学文献，包括把工程化生物学理解为编程的写作，则覆盖了书写生物代码这一前沿。

关于最深层的汇合——生物学的信息论理解、进化与生命的计算视角——跨领域研究者的工作是前沿所在，包括系统生物学文献，以及关于生命系统热力学与信息论的研究。这些内容主要存在于一手文献中。

资源 作用 类型
Jumper 等，AlphaFold 论文（2021，免费） 蛋白质结构预测的里程碑 深入
Mitchell，An Introduction to Genetic Algorithms 进化计算入口 入门
Adleman，DNA 计算论文（1994，免费） DNA 计算的源头 深入
机器学习在生物学中的文献（当前） 活跃前沿 参考
AlphaFold Protein Structure Database（免费） 大规模开放蛋白质结构预测 实践
EMBL-EBI AlphaFold practical guide（免费） 使用 AlphaFold 和理解置信度的实践指南 实践
Rosalind 平台（免费） 通过问题学习生物信息学 实践

陷阱

陷阱 为什么会误导 更好的回应
把“DNA 是代码”看成纯粹隐喻 把 DNA 描述为代码、把细胞描述为计算机，听起来可能像松散类比，好像只是借用计算机术语谈论生物学。但它不只是这样：DNA 确实是严格意义上的数字代码，也就是由字母表上的序列构成，并由分子机器读取和翻译；生命的信息处理视角是描述性的，不只是隐喻。把它 dismiss 成隐喻，会错过关于生命最深的事实之一。 在它确实字面成立的地方，认真对待生命的信息处理视角。DNA 是数字代码；中心法则描述信息流；细胞通过分子网络处理信息。计算概念描述的是生命系统实际在做什么，而不是借来的比喻。同时，也要保持适当谨慎，区分哪些地方只是类比延伸：细胞并不是冯·诺伊曼计算机。应区分生命的信息性质这一字面事实，与更松散的计算类比。
夸大从生物到 CS 的启发作用 受生物启发的方法，例如进化计算、神经网络，有时会被宣传成更优越，只因为它们来自生物学，好像生物启发本身就保证成功。实际上，生物启发是想法来源，不是质量保证；最成功的受生物启发方法——神经网络——正是通过大幅偏离其生物源头而成功的。生物启发是一种产生方法的启发式来源，而不是对方法的验证。 根据方法本身的表现来评估受生物启发的方法，而不是根据其生物血统来评估。进化计算适用于特定问题，例如复杂、不可微的搜索空间，但不是通用的优越方法；神经网络的成功来自放弃生物忠实度，转向真正有效的东西。生物启发说明这些想法从哪里来；它们是否有效，是另一个问题，只能由表现回答，而不是由起源回答。
低估计算生物学已经变成什么 受过计算机科学训练的人，可能会把生物学看作一个遥远领域，认为它和自己的技能关系不大。这种看法越来越错误：生物学在很大程度上已经成为一门计算科学，计算技能是现代生物学研究的核心，尤其体现在基因组学、结构生物学、系统生物学和药物发现中。计算背景的人如果把生物学视为无关领域，就会错过一个最大、也最具后果的技能应用场域。 把生物学理解为一个计算技能越来越不可或缺的领域，而且其中的问题具有极高后果。生物学转变为数据与计算密集型科学，意味着受过计算训练的人将在生物学研究中扮演越来越重要的角色。对于正在决定将计算技能应用到哪里的学习者来说，计算与生物学的汇合，是最重要、最有影响力的前沿之一。