27 98 32 120 0 905 36 23 0 0 89 246 103 134 0 198 1 148 1153 0 716 240 9 139 125 11 28 81 23 240 535 86 28 606 43 10 65 64 77 24 44 18 61 0 7 41 15 34 0 0 73 11 7 44 257 26 464 318 71 0 153 83 27 26 46 18 72 90 1 0 0 114 30 17 0 336 527 243 18 14 14 0 0 0 0 15 48 196 157 0 92 250 103 42 13 19 153 51 34 94 12 32 33 17 11 409 154 495 95 161 56 79 234 35 24 17 96 62 46 245 371 26 229 66 16 53 34 30 22 192 33 136 104 13 78 550 0 201 23 0 0 0 0 0 27 0 46 0 0 76 0 75 0 24 8 95 0 96 0 22 0 127 37 28 13 0 698 0 34 42 61 208 24 15 18 49 35 37 54 44 889 175 10 258 12 48 30 157 0 28 0.087127 0.040904 0.040432 0.046872 0.033474 0.038255 0.049530 0.088612 0.033618 0.036886 0.085357 0.080482 0.014753 0.039772 0.050680 0.069577 0.058542 0.010494 0.029916 0.064718 Ala Arg Asn Asp Cys Gln Glu Gly His Ile Leu Lys Met Phe Pro Ser Thr Trp Tyr Val S_ij = S_ji and PI_i for the Dayhoff model, with the rate Q_ij=S_ij*PI_j The rest of the file is not used. Prepared by Z. Yang, March 1995. See the following reference for notation used here: Yang, Z., R. Nielsen and M. Hasegawa. 1998. Models of amino acid substitution and applications to mitochondrial protein evolution. Mol. Biol. Evol. 15:1600-1611. ----------------------------------------------------------------------- 30 109 17 154 0 532 33 10 0 0 93 120 50 76 0 266 0 94 831 0 422 579 10 156 162 10 30 112 21 103 226 43 10 243 23 10 66 30 36 13 17 8 35 0 3 95 17 37 0 0 75 15 17 40 253 57 477 322 85 0 147 104 60 23 43 39 29 17 0 0 0 20 7 7 0 57 207 90 20 7 7 0 0 0 0 17 20 90 167 0 17 345 67 27 10 10 93 40 49 50 7 43 43 4 7 772 137 432 98 117 47 86 450 26 20 32 168 20 40 269 590 20 169 57 10 37 31 50 14 129 52 200 28 10 73 696 0 27 3 0 0 0 0 0 3 0 13 0 0 10 0 17 0 20 3 36 0 30 0 10 0 40 13 23 10 0 260 0 22 23 6 365 20 13 17 33 27 37 97 30 661 303 17 77 10 50 43 186 0 17 A R N D C Q E G H I L K M F P S T W Y V Ala Arg Asn Asp Cys Gln Glu Gly His Ile Leu Lys Met Phe Pro Ser Thr Trp Tyr Val Accepted point mutations (x10) Figure 80 (Dayhoff 1978) ------------------------------------------------------- A 100 /* Ala */ A 0.087 /* Ala */ R 65 /* Arg */ R 0.041 /* Arg */ N 134 /* Asn */ N 0.040 /* Asn */ D 106 /* Asp */ D 0.047 /* Asp */ C 20 /* Cys */ C 0.033 /* Cys */ Q 93 /* Gln */ Q 0.038 /* Gln */ E 102 /* Glu */ E 0.050 /* Glu */ G 49 /* Gly */ G 0.089 /* Gly */ H 66 /* His */ H 0.034 /* His */ I 96 /* Ile */ I 0.037 /* Ile */ L 40 /* Leu */ L 0.085 /* Leu */ K 56 /* Lys */ K 0.081 /* Lys */ M 94 /* Met */ M 0.015 /* Met */ F 41 /* Phe */ F 0.040 /* Phe */ P 56 /* Pro */ P 0.051 /* Pro */ S 120 /* Ser */ S 0.070 /* Ser */ T 97 /* Thr */ T 0.058 /* Thr */ W 18 /* Trp */ W 0.010 /* Trp */ Y 41 /* Tyr */ Y 0.030 /* Tyr */ V 74 /* Val */ V 0.065 /* Val */ scale factor = SUM_OF_PRODUCT = 75.246 Relative Mutability The equilibrium freqs. (Table 21) Table 22 (Dayhoff 1978) Dayhoff (1978) ---------------------------------------------------------------- Some notes from 1995, for those technical people: I managed to find some notes I wrote in 1995. The symbols are not that comprehensible now, but you can get the basic idea, I think. (1) Construction of P(0.01), for 1 PAM p_ij(0.01) = m_i * A_{ij}/\sum_k{A_{ik}} / 7524.6 (2) Eigensolution of P(0.01) = exp{Q*0.01} P(0.01) = U diag{\lambda...} U^{-1} Then Q = U diag{100*log{\lambda}...} U^{-1} I did not use the PAM transition probabilities as rates assuming 0.01 is close to 0, but instead take them as P(0.01) to recover the rate matrix, and as we expect, the rates are more different from each other than the p_ij(0.01) are. I seem to recall that I thought some details in the Dayhoff paper and the Kishino et al. (1990) paper were not entirely right. I think I thought that Q should be a symmetrical matrix, right-multiplied by a diagonal matrix, while either Dayhoff or Kishino or both used left-multiplication. As far as I know, codeml and protml give very similar (but not identical, I think) results under the Dayhoff model. My jones.dat file is not based on the Jones et al. (1992) paper, but is based on an updated data set sent to me by David Jones. So codeml and protml gave different results under JTT, but ranking of trees was not affected for the data set I tested. Ziheng Yang