This preview shows page 1. Sign up to view the full content.
Unformatted text preview: ill defeated by exponential explosion.
¶8. Lipton (1995) estimates that it is feasible for n 70, but this is also
within the range of conventional computer.
¶9. Massive parallelism: Nevertheless, Adleman’s algorithm illustrates
the massive parallelism of molecular computation.
Step 1 (generation of all possible paths) took about an hour for n = 7.
Adleman estimates that about 1014 ligation operations were performed,
and that it could be scaled up to 1020 operations.
Therefore, speeds of about 1015 to 1016 ops/sec (1–10 petaoperations/s)
should be achievable.
That is, digital supercomputer range. 236 CHAPTER IV. MOLECULAR COMPUTATION ¶10. Energy: Adlemen estimates that 2 ⇥ 1019 ligation operations were
performed per joule of energy.
Contemporary supercomputer perform only 109 operations per joule,
so MC is 1010 more energy e cient.
¶11. It is near the thermodynamic limit of 34 ⇥ 1019 operations per joule.
Recall (Ch. II, Sec. B.1) kT ln 2 ⇡ 3 ⇥ 10 9 pJ = 3 ⇥ 10 21 J, so there
can be about 3.3 ⇥ 1020 bit changes/J.
¶12. Space: One bit of information occupies about 1 cubic nm.
Contemporary disks store a bit in about 1010 cubic nm.
That is, a 1010 improvement in density.
¶13. Inherent error: A more pervasive problem is the inherent error in
the ﬁltering processes (due to incorrect hybridization).
Some strands we don’t want, get through; and some that we do want,
don’t.
With many ﬁltering stages the errors accumulate to the extent that the
algorithms fail.
¶14. There are some approaches to errorresistant DNA computing....
View Full
Document
 Fall '13
 BruceMacLennan

Click to edit the document details