107 
DR. GELFAND: We are contemplating trying to develop alternate pro- 
karyotic systems — for example, Streptomyces , soil organisms. For a great 
many potential experiments, E. col i K-12 is totally inappropriate, not 
because it is unsafe, but because the DNA might not be expressed in it, 
because E. col i K-12 isn't known to produce antibiotics. One might want 
to attempt to use these techniques to develop new and safer antibiotics. 
DR. FREDRICKSON: Are there any other comments from the Committee, 
or questions? 
There being none, we will move to the public witnesses in this section. 
The three public witnesses who have asked for time include three who have 
spoken several times previous to this. I must keep us to our schedule as 
closely as possible. I will therefore allot you three minutes, each of you, 
beginning with Dr. Schwartz. 
DR. SCHWARTZ: I wonder if I can protest that. I have cut out half 
of my talk to get it down to five. It is very difficult, and it is fairly 
technical . 
DR. FREDRICKSON: Then I would suggest, Dr. Schwartz, that you speak; 
and perhaps we will ask several of the others to wait until this evening when 
we will have makeup time. You may have five minutes. 
DR. SCHWARTZ: 
I will try to get through it. 
It has been asserted that the security of biological containment is 
evidenced by exceptionally low probabilities of organisms escaping, surviving, 
and ultimately causing infection. I submit that the assumptions and calcu- 
lations leading to these optimistically low probabilities are not valid. The 
reliability theory has had its greatest success in dealing with electronic 
systems. Its use in assessing the reliability of a particular system requires 
knowledge of all significant component failures and failure combinations that 
are possible in that system. It is necessary to test the entire system to 
determine the individual rates of component failure. 
In addition, the choice of the appropriate mathematical model depends 
upon observation of system performance. The laboratory experiments on 
isolated components do not include environmental effects, nor do they take 
into account interactions between various components. Included in the system 
we are considering here are not just the vector, host, and foreign DNA, but 
also the researchers, maintenance workers, students, laboratory animals, 
and tests. Perhaps we must also consider the possibility of unauthorized 
intruders as well. 
The accepted risk analysis has implicitly assumed that separate com- 
ponents of a system act independently, and that therefore one may compute the 
probability of their simultaneous failure by multiplying the probabilities of 
the individual failures. This is a serious error. 
[ 311 ] 
