Nanomedicine, Volume I: Basic Capabilities
© 1999 Robert A. Freitas Jr. All Rights Reserved.
Robert A. Freitas Jr., Nanomedicine, Volume I: Basic Capabilities, Landes Bioscience, Georgetown, TX, 1999
Many important medical nanorobotic tasks will require computation during the acquisition and processing of sensor data, the control of tools, manipulators, and motility systems, navigation and communication, and during the coordination of collective activities with neighboring nanorobots. Ex vivo computation has few theoretical limits, but computation by in vivo nanorobots will be subject to a number of constraints such as physical size, power consumption, onboard memory and processing speed.
The memory required onboard a medical nanorobot will be strongly mission dependent. Recognizing and manipulating molecules is fundamental. A very simple mission might demand only the identification or handling of perhaps ~10 different molecules. For example, basic respiratory gas transport nanorobots such as respirocytes1400 (Chapter 22) may require the operation of fixed-shape receptors that bind simple molecules such as O2, CO2, H2O, and glucose. Identifiers for such receptors may need only a few bits, hence this memory requirement should be negligible. A toxin removal device (Chapter 19) similarly may require keeping track of only a few types of fixed-shape receptors. On the other hand, a survey or assay mission might need to recognize N = 100-1000 distinct proteins. A spherical 1-micron nanorobot can have >104 fixed-shape receptors on its surface; if these will suffice, then we require N log2 (N) ~ 104 bits to identify each of N = 1000 different receptor types. This seems more efficient than using more advanced reconfigurable receptors (Section 18.104.22.168), which might need >104 bits per receptor-pattern to specify each binding site geometry to the necessary atomic-scale resolution (Section 22.214.171.124), thus imposing a total memory requirement of >107 bits for an onboard library of N = 1000 different receptor types. Consequently, simple missions involving basic process control with limited motility may require no more than ~105-106 bits of memory, comparable to an old Apple II computer (including RAM plus floppy disk drive). At the other extreme, a complex cell repair mission might require the onboard storage of a substantial fraction of the patient's genetic code, representing ~109 bits of memory including perhaps ~0.2 x 109 bits of linear sequence data for all 100,000 protein types found in the human body, again assuming 300 amino acids per protein. (Most amino acids are folded into the protein's interior and are not readily accessible to surface probing unless the protein is unfolded, which usually is not desirable or convenient. On the other hand, binding sites for large molecules should be physically easier to construct than small-molecule receptors; Section 3.5.9.) An onboard memory of 109-1010 bits would be in the same range as the 1985 Cray-2 (2 x 1010 bits) or the 1989 Cray-3 (6 x 108 bits) supercomputers.1
Computational speed will also be strongly mission dependent. However, extremely simple process control systems in basic factory settings may only require speeds as slow as 104 bit/sec (Chapter 12). Individual natural biocomputational devices (as opposed to multiple such biodevices operating in parallel) generally do not exceed this speed. Examples include mRNA translation during protein manufacture at ~15 Hz (~75 bits/sec assuming 5 bits/protein);997 transcription from DNA by RNA polymerase at ~40 Hz (~80 bits/sec at 2 bits/nucleotide);997 DNA replication at ~800 Hz (~1600 bits/sec at 2 bits/nucleotide);997 typically 5-100 Hz (bits/sec) for neural electrical discharges (Section 4.8.6); ~1000 Hz (bits/sec) for excitory cholinergic synapses, and gated ion channels at ~104 Hz (bits/sec) (Section 126.96.36.199). At the other extreme, a processing speed of 109 bits/sec allows a ~109 bit genomic information store to be processed in ~1 sec, the small-molecule diffusion time across an average 20-micron wide cell. In 1998, personal desktop computers capable of ~109-1010 bits/sec (~108 operations/sec) were commonly available.
This Section describes possible nanomechanical (Section 10.2.1) and nanoelectronic (Section 10.2.2) computers, biocomputers (Section 10.2.3), and briefly examines the ultimate limits to computation including reversible and quantum computing (Section 10.2.4). Computer architectural issues are not explicitly addressed here. (For example, advanced architectures, especially large CPU systems, may employ distributed clocks and a distributed computing network.)
Last updated on 23 February 2003