PFRMAT AL 
TARGET T0077 
AUTHOR 5827-4749-3439 
METHOD T0077 structure prediction 
METHOD 
METHOD As usual, we employed a two-pronged approach: (1) We scored the  
METHOD target against all the HMMs and threading models in our libraries,  
METHOD and (2) we gathered homologs to the target using iterated FASTA 
METHOD (where new hits are used as query sequences until convergence),  
METHOD constructed an HMM for the target and homologs  
METHOD (using UCSC's HMM software), and scored PDB.  
METHOD This yielded three sets of scores, which we then used  
METHOD to find a target-structure match. 
METHOD 
METHOD In this prediction, all scores were rather weak. This target-structure 
METHOD match satisfied all criteria better than other matches did.  
METHOD 
METHOD One of the target homologs scored d1pkp_1 as the top threading 
METHOD prediction, and this score was higher than any other target homolog 
METHOD with any other structure.  Nevertheless, d1pkp_1 did not appear in 
METHOD the top 10 threading hits for a number of other target homolog  
METHOD sequences. 
METHOD >gi|133977|sp|P12743|RS6X_HALMA 
METHOD EALEVARDTGAVKKGTNETTKSIERGSAELVFVAEDVQPEEIV--MHIPELADEKGVPFIF 
METHOD VEQQDDLGHAAGLEVGSAAAAVTDAG 
METHOD >d1pkp_1 
METHOD -------------GTTIPHEVIGHFGAGEIILKPAS-EGTGVIAGGPARAVLELAGISDIL 
METHOD SKSIGSNTPINMVRATFDGLKQLK-- 
METHOD 
METHOD Despite the overall weak scores, we decided to make a prediction 
METHOD since both threading and HMMs ranked d1pkp_1 among the top folds. 
METHOD  
METHOD The alignment submitted was created by aligning the target and structure  
METHOD to an HMM constructed to represent both the target and the structure, 
METHOD as follows: (1) Phylogenetic analysis (using Sjolander's Bete method) 
METHOD of the target and homologs produced ten subfamilies.  
METHOD (2) For each of these subfamilies, we constructed an HMM 
METHOD and scored all the structure homologs. One of the subfamilies gave the 
METHOD highest probability to one of the structure homologs.  
METHOD (3) This subfamily HMM was reestimated using the structure, the 
METHOD highest-scoring structure homolog, the target, and the target homologs 
METHOD having the highest affinity in both threading and HMMs for the 
METHOD structure, as training sequences. In UCSC SAM parlance, this was 
METHOD accomplished by using the buildmodel program, employing 
METHOD Dirichlet mixture priors, setting initial_noise and anneal_noise 
METHOD to zero, giving the subfamily HMM as the initial model, and 
METHOD setting the training sequences to the 5 sequences chosen. 
METHOD The alignment was obtained by the Viterbi algorithm 
METHOD (align2model), as usual. 
METHOD  
METHOD This alignment was examined (but not edited in any way) 
METHOD to check that the target and structure retained their 
METHOD respective alignments with their own homologs. 
METHOD Four of the sequences are shown aligned below. 
METHOD 
METHOD 
METHOD  
METHOD                                        10        20        30 
METHOD                                         |         |         | 
METHOD  1pkp_1                         XSSTEARXTHERMXPHILXS-----GTTIPHEVIGHFG 
METHOD gi|1707008                      TFPHRSEGDYG-AAKVMLRPASPGTG------------ 
METHOD 
METHOD gi|2500373|sp|Q53602|YBXF_STAAU -------------------------GLKETLKALKKDQ 
METHOD T0077                           MAPVKSQESINQKLALVIKSGKYTLGYKSTVKSLRQGK 
METHOD 
METHOD 
METHOD                                 40        50        60        70 
METHOD                                  |         |         |         | 
METHOD d1pkp_1                         AGEIILKPASEGTG-----VIAGGPARAVLELAGISDI 
METHOD gi|1707008                      -------------------VIAGGAVRIVLEMAGVENA 
METHOD 
METHOD gi|2500373|sp|Q53602|YBXF_STAAU VTSLIIAEDVEVYLMTRVLSQINQ-KNIPVSFFKSKHA 
METHOD T0077                           SKLIIIAANTPVLRKSELEYYAMLSKTKVYYFQGGNNE 
METHOD 
METHOD 
METHOD                                   80         90       100 
METHOD                                   |          |         | 
METHOD d1pkp_1                         LSKSIGSN.TPINMVR----ATFDGLKQLK 
METHOD gi|1707008                      LGKQLGSN.NALNNAR----ATLAAVQQMR 
METHOD  
METHOD gi|2500373|sp|Q53602|YBXF_STAAU LGKHVGINvNATIVAL-------------- 
METHOD T0077                           LGTAVGKL.FRVGVVSILEAGDSDILTTLA 
METHOD  
METHOD 
MODEL 1 
PARENT 1pkp 
V    17    I    81 
I    18    P    82 
K    19    H    83 
S    20    E    84 
G    21    V    85 
K    22    I    86 
Y    23    G    87 
T    24    H    88 
L    25    F    89 
G    26    G    90 
Y    27    A    91 
K    28    G    92 
S    29    E    93 
T    30    I    94 
V    31    I    95 
L    34    L    96 
R    35    K    97 
Q    36    P    98 
G    37    A    99 
K    38    S    100 
S    39    E    101 
K    40    G    102 
L    41    T    103 
I    42    G    104 
I    43    V    105 
I    44    I    106 
A    45    A    107 
A    46    G    108 
N    47    G    109 
T    48    P    110 
P    49    A    111 
V    50    R    112 
K    66    A    113 
V    67    V    114 
Y    68    L    115 
Y    69    E    116 
F    70    L    117 
Q    71    A    118 
G    72    G    119 
G    73    I    120 
N    74    S    121 
N    75    D    122 
E    76    I    123 
L    77    L    124 
G    78    S    125 
T    79    K    126 
A    80    S    127 
V    81    I    128 
G    82    G    129 
K    83    S    130 
L    84    N    131 
F    85    T    132 
R    86    P    133 
V    87    I    134 
G    88    N    135 
V    89    M    136 
V    90    V    137 
S    91    R    138 
I    92    A    139 
D    97    T    140 
S    98    F    141 
D    99    D    142 
I    100    G    143 
L    101    L    144 
T    102    K    145 
T    103    Q    146 
L    104    L    147 
A    105    K    148 
TER 
END 
