June 10, 1955
Dear Dr. Demerec:

Thank you for sending the tabulations on your 4-point linkage test on
tryD-10 cys Bel2 x— try A& me Alég.

We were most interatted to go over it-—- Dr, Morse and I particularly in
some detail. After laborfiously testing all the sequencies "geometrically"
I worked out a "matrix" test in which this has been done once, and one can
quickly analyze further batches of data. I anclose an expianation with some
illustrations of its use; I am not suse how intelligible this will be » but
once it is set up it saves a tremendous anount of time and mistakes.

The 3~point data you sent me went in fairly obvious fashion: I can easily
understanl the premises of your argument. The 4-point test unfortunately shows
some discrepancies in regard to meth. In the En(cys,meth) for exa ple,
you show:

cys 362 cys-meth 365 cys-anth 1453 orig. cys-meth-(anth) 1453.
There seams to be aadiscrepancy as to whether there are 2 cya7meth™ or not.

Similarly, in the En(cys,meth,anth), there is cys 395 cys meth 397 but
cys—anth 1635 cys-meth-anth 1635.

Was there a simple typographical error? These confused instances are the only
suggestion of M” types, which may throw some doubt on the linkage here, but to
illustrate the use of the matrix method, I assumed the validity of cys~meth™ as
a rare class.
not all the

In any event, largely because of the epistatis of tryD™” over tryAS, thaxunty
sequences are excluded, and further experiments would be needed. Do my deductions
correspond to your own.

My only general comment is that the analysis Sives relatively little room
to check internal consistency. I should think it would be extremely important to
concentrate, for example, on a valid 4~point series, and to test this in all
possible ways, not only by mute reciprocal transductions, but also by using dif-
ferent combinations of the markers as can be obtained from the transductions then-
selves. Otherwise, ons can conclude that if the linear sequence model is correct,
and there are no complications, then only one sequence will fit, but there will
have been too little assurance that the sequence is in fact linear and unique.

Have you paid any attention to quantitative mapping? The date from a single
transduction seem reasonably consistent, but I have not tried to handle the figures
except as many/few, especially as I thought there was some question about part of
the data. If you will reassure me about these, moo or want to discuss any other,

I will be delighted to work on same.
Yours sincerely,

G (4 #7 | Joshua Lederberg
turgor welata, 82