Abstract:

The paper presented here is Concerned with the numerical cladisties. In

consideration of the fact that the parallel evolution has close relation to the length

of evolution graph, a new method of reconstructing evolutionary tree has been de-

veloped for the application and practice of cladistics.

     The procedure of the algorithm of the new method presented in Table I is

similar to the method described in paper "An algorithm for cladistics method

of maximal same step length".

     An essential step of the algorithm is how to decide the coefficient between two

cladistic units (CTUs). A coefficient called parallel evolutionary coefficient between

CTUp and CTUq is defined as follows:

                                                                

where the j is code of CTU and the i is code of character;  E(p, q, i, j) is a func-

tion given by following expression:

min (X_ij, Xpj)+(X_ij, Xqj)-2min(Xpj, Xqj) as Xij>min (Xpj, Xqj)

E(p,q, i,j ) =

                 0                                               otherwise.

where the X_ij is the ith row (CTU) jth colunm (Character) element of the data

matrix.

      Because the method of minimal parallel evolution is closely related to the

length of evolutionary graph, it is superior to the method of maximal same step

length. A simple datum as an example for comparison shows that the method of

minimal parallel evolution can arrive at a better result.

    But in some cases, we may combine one method with another and thus the

coefficient should take following form:

                    S(S_ij)=M·S (C) _ij-N·S(P) _ij

in which S (C) _ij and S (P) _ij are the same step coefficients and the parallel evolu-

tion coefficient respectively, and the M and N are positive integers as a weight

number being given in advance.

Key words: Cladistics, Method of minimal parallel evolution