Full model selection results:
Cross validation was done on the positive acceptor model for the number of states (bw%i)
of the full connected HMM behind the splice site.
The first part is modelled via a 'linear' HMM (just observation frequencies).
The accuracy (c:%f), error rate (w:%f), false positives (fp:%f), true positives (tp:%f) is shown for the
connected model (linear+full connected) on the validation set (only for the positive model).
The optimal bias term (thresh*:%f) for the validation data is shown. This treshhold was used on the test set.
The model with highest accuracy was chosen.

ag_pos_bw10_tr100_r1.out:of 5000 samples (c:0.836800,w:0.163200,fp:0.159178,tp:0.830149,tresh*:-87.71853694492961)
ag_pos_bw10_tr100_r2.out:of 5000 samples (c:0.869400,w:0.130600,fp:0.121202,tp:0.854145,tresh*:-87.246716862678582)
ag_pos_bw10_tr100_r3.out:of 5000 samples (c:0.902400,w:0.097600,fp:0.091444,tp:0.892673,tresh*:-66.669328433937153)
ag_pos_bw10_tr100_r4.out:of 5000 samples (c:0.629400,w:0.370600,fp:0.108949,tp:0.208246,tresh*:-2065.3603964794147)
ag_pos_bw10_tr100_r5.out:of 4999 samples (c:0.872374,w:0.127626,fp:0.105871,tp:0.836344,tresh*:-65.455810536879113)
ag_pos_bw11_tr100_r1.out:of 5000 samples (c:0.833200,w:0.166800,fp:0.160141,tp:0.822187,tresh*:-88.224252001242519)
ag_pos_bw11_tr100_r2.out:of 5000 samples (c:0.871000,w:0.129000,fp:0.132515,tp:0.876705,tresh*:-87.467456220306502)
ag_pos_bw11_tr100_r3.out:of 5000 samples (c:0.896400,w:0.103600,fp:0.096342,tp:0.884933,tresh*:-66.777892976343992)
ag_pos_bw11_tr100_r4.out:of 5000 samples (c:0.894800,w:0.105200,fp:0.082685,tp:0.858559,tresh*:-66.777149272489339)
ag_pos_bw11_tr100_r5.out:of 4999 samples (c:0.852370,w:0.147630,fp:0.086622,tp:0.751328,tresh*:-65.336829907365086)
ag_pos_bw12_tr100_r1.out:of 5000 samples (c:0.832200,w:0.167800,fp:0.165597,tp:0.828556,tresh*:-88.203619401972105)
ag_pos_bw12_tr100_r2.out:of 5000 samples (c:0.866600,w:0.133400,fp:0.124758,tp:0.852571,tresh*:-87.641910314357517)
ag_pos_bw12_tr100_r3.out:of 5000 samples (c:0.833800,w:0.166200,fp:0.084585,tp:0.704850,tresh*:-66.491506616460455)
ag_pos_bw12_tr100_r4.out:of 5000 samples (c:0.892800,w:0.107200,fp:0.101816,tp:0.884134,tresh*:-68.161988885962103)
ag_pos_bw12_tr100_r5.out:of 4999 samples (c:0.645329,w:0.354671,fp:0.124479,tp:0.264081,tresh*:-2062.8550802051891)
ag_pos_bw13_tr100_r1.out:of 5000 samples (c:0.829000,w:0.171000,fp:0.149551,tp:0.793524,tresh*:-87.491164336612087)
ag_pos_bw13_tr100_r2.out:of 5000 samples (c:0.871200,w:0.128800,fp:0.096315,tp:0.818468,tresh*:-86.014965043797105)
ag_pos_bw13_tr100_r3.out:of 5000 samples (c:0.899600,w:0.100400,fp:0.099608,tp:0.898349,tresh*:-67.918239057959624)
ag_pos_bw13_tr100_r4.out:of 5000 samples (c:0.886800,w:0.113200,fp:0.107328,tp:0.877349,tresh*:-68.341075019446009)
ag_pos_bw13_tr100_r5.out:of 4999 samples (c:0.645329,w:0.354671,fp:0.124479,tp:0.264081,tresh*:-2062.8550802051891)
ag_pos_bw14_tr100_r1.out:of 5000 samples (c:0.817600,w:0.182400,fp:0.134467,tp:0.738323,tresh*:-86.966198352467543)
ag_pos_bw14_tr100_r2.out:of 5000 samples (c:0.867800,w:0.132200,fp:0.117647,tp:0.844176,tresh*:-88.356749500112414)
ag_pos_bw14_tr100_r3.out:of 5000 samples (c:0.893000,w:0.107000,fp:0.097322,tp:0.877709,tresh*:-68.266626936094752)
ag_pos_bw14_tr100_r4.out:of 5000 samples (c:0.895400,w:0.104600,fp:0.092412,tp:0.875783,tresh*:-67.584608693674951)
ag_pos_bw14_tr100_r5.out:of 4999 samples (c:0.827165,w:0.172835,fp:0.085659,tp:0.682784,tresh*:-66.034174406973278)
ag_pos_bw5_tr100_r1.out:of 5000 samples (c:0.837200,w:0.162800,fp:0.158537,tp:0.830149,tresh*:-87.010297633495711)
ag_pos_bw5_tr100_r2.out:of 5000 samples (c:0.877000,w:0.123000,fp:0.139948,tp:0.904512,tresh*:-87.821148957883025)
ag_pos_bw5_tr100_r3.out:of 5000 samples (c:0.899800,w:0.100200,fp:0.094056,tp:0.890093,tresh*:-66.391166340644915)
ag_pos_bw5_tr100_r4.out:of 5000 samples (c:0.883400,w:0.116600,fp:0.082361,tp:0.828288,tresh*:-66.799304370542046)
ag_pos_bw5_tr100_r5.out:of 4999 samples (c:0.868174,w:0.131826,fp:0.064806,tp:0.757173,tresh*:-64.125346514658332)
ag_pos_bw6_tr100_r1.out:of 5000 samples (c:0.838400,w:0.161600,fp:0.138318,tp:0.799894,tresh*:-85.964902664929284)
ag_pos_bw6_tr100_r2.out:of 5000 samples (c:0.878400,w:0.121600,fp:0.147059,tp:0.919727,tresh*:-88.079393099171)
ag_pos_bw6_tr100_r3.out:of 5000 samples (c:0.901200,w:0.098800,fp:0.101568,tp:0.905573,tresh*:-66.593561173111539)
ag_pos_bw6_tr100_r4.out:of 5000 samples (c:0.893000,w:0.107000,fp:0.093061,tp:0.870564,tresh*:-66.856669292010423)
ag_pos_bw6_tr100_r5.out:of 4999 samples (c:0.869374,w:0.130626,fp:0.084376,tp:0.792774,tresh*:-64.382079617093069)
ag_pos_bw7_tr100_r1.out:of 5000 samples (c:0.838600,w:0.161400,fp:0.152118,tp:0.823248,tresh*:-86.819085148163396)
ag_pos_bw7_tr100_r2.out:of 5000 samples (c:0.874800,w:0.125200,fp:0.130575,tp:0.883526,tresh*:-87.508311098573216)
ag_pos_bw7_tr100_r3.out:of 5000 samples (c:0.897200,w:0.102800,fp:0.093730,tp:0.882869,tresh*:-66.656119865171021)
ag_pos_bw7_tr100_r4.out:of 5000 samples (c:0.897000,w:0.103000,fp:0.080739,tp:0.861169,tresh*:-66.359143040873136)
ag_pos_bw7_tr100_r5.out:of 4999 samples (c:0.890378,w:0.109622,fp:0.095284,tp:0.866631,tresh*:-64.839617152873274)
ag_pos_bw8_tr100_r1.out:of 5000 samples (c:0.834800,w:0.165200,fp:0.166560,tp:0.837049,tresh*:-87.678284952465049)
ag_pos_bw8_tr100_r2.out:of 5000 samples (c:0.874000,w:0.126000,fp:0.136716,tp:0.891396,tresh*:-87.924963703335237)
ag_pos_bw8_tr100_r3.out:of 5000 samples (c:0.901600,w:0.098400,fp:0.095363,tp:0.896801,tresh*:-66.221422354140884)
ag_pos_bw8_tr100_r4.out:of 5000 samples (c:0.897200,w:0.102800,fp:0.076524,tp:0.854906,tresh*:-67.268692569542281)
ag_pos_bw8_tr100_r5.out:of 4999 samples (c:0.824765,w:0.175235,fp:0.068014,tp:0.647184,tresh*:-64.503085554447665)
ag_pos_bw9_tr100_r1.out:of 5000 samples (c:0.838800,w:0.161200,fp:0.161746,tp:0.839703,tresh*:-87.580949364588975)
ag_pos_bw9_tr100_r2.out:of 5000 samples (c:0.871200,w:0.128800,fp:0.129929,tp:0.873033,tresh*:-87.07675545460944)
ag_pos_bw9_tr100_r3.out:of 5000 samples (c:0.896600,w:0.103400,fp:0.102874,tp:0.895769,tresh*:-67.283882714419761)
ag_pos_bw9_tr100_r4.out:of 5000 samples (c:0.901800,w:0.098200,fp:0.078470,tp:0.870042,tresh*:-66.73439069614642)
ag_pos_bw9_tr100_r5.out:of 4999 samples (c:0.820364,w:0.179636,fp:0.053898,tp:0.612115,tresh*:-63.998041426936481)

Best models:
ag_pos_bw9_tr100_r1.out:of 5000 samples (c:0.838800,w:0.161200,fp:0.161746,tp:0.839703,tresh*:-87.580949364588975)
ag_pos_bw6_tr100_r2.out:of 5000 samples (c:0.878400,w:0.121600,fp:0.147059,tp:0.919727,tresh*:-88.079393099171)
ag_pos_bw10_tr100_r3.out:of 5000 samples (c:0.902400,w:0.097600,fp:0.091444,tp:0.892673,tresh*:-66.669328433937153)
ag_pos_bw9_tr100_r4.out:of 5000 samples (c:0.901800,w:0.098200,fp:0.078470,tp:0.870042,tresh*:-66.73439069614642)
ag_pos_bw7_tr100_r5.out:of 4999 samples (c:0.890378,w:0.109622,fp:0.095284,tp:0.866631,tresh*:-64.839617152873274)

Performance of the best joined (positive with corresponding negative) models on the validation data with optimal treshhold:

1) error rate

a) for all 5 runs
0.1570 0.1130 0.0662 0.0744 0.0834

b) mean
0.0988

c) standard deviation
0.0370

2) specificity ( 1 - false positives )

a) for all 5 runs
0.8399 0.8662 0.9422 0.9280 0.9230

b) mean
0.8999

c) standard deviation
0.0443

3) sensitivity (true positives)

a) for all 5 runs
0.8482 0.9208 0.9205 0.9217 0.9060

b) mean
0.9034

c) standard deviation
0.0316