Dose-Response Curve Fitter

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416

<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Dose-Response Curve Fitter</title>
<link rel="icon" href="data:,">
<style>
/* sloprun design tokens — inline this block into every demo (self-contained rule).
   Identity: instrument-panel. Machine facts in mono; human words in sans.
   Green is EARNED: only for "it ran / it worked" states, never decoration. */
:root {
  --bg: #F6F6F3;          /* neutral paper, not cream */
  --surface: #FFFFFF;
  --ink: #1A1D21;
  --muted: #5A6068;
  --line: #E3E2DC;
  --accent: #C05600;      /* phosphor amber, darkened for light bg */
  --accent-soft: #FFF3E6;
  --run: #1A7F37;         /* earned green */
  --run-soft: #E7F4EA;
  --danger: #C93C37;
  --radius: 10px;
  --font-sans: system-ui, -apple-system, "Segoe UI", sans-serif;
  --font-mono: ui-monospace, "SF Mono", "Cascadia Code", Menlo, Consolas, monospace;
}
@media (prefers-color-scheme: dark) {
  :root:not([data-theme="light"]) {
    --bg: #14161A; --surface: #1C1F26; --ink: #E8E6E1; --muted: #9AA0A8;
    --line: #2A2E36; --accent: #FFB454; --accent-soft: #2A2118;
    --run: #3FB950; --run-soft: #16281B; --danger: #F47067;
  }
}
:root[data-theme="dark"] {
  --bg: #14161A; --surface: #1C1F26; --ink: #E8E6E1; --muted: #9AA0A8;
  --line: #2A2E36; --accent: #FFB454; --accent-soft: #2A2118;
  --run: #3FB950; --run-soft: #16281B; --danger: #F47067;
}
body { background: var(--bg); color: var(--ink); font-family: var(--font-sans); }

/* ---------- base ---------- */
* { box-sizing: border-box; }
[hidden] { display: none !important; }
html, body { margin: 0; padding: 0; }
body { -webkit-text-size-adjust: 100%; line-height: 1.5; overflow-x: hidden; }
.wrap { max-width: 1120px; margin: 0 auto; padding: 22px 16px 28px; }

h1, h2, h3 { margin: 0; font-weight: 650; letter-spacing: -0.01em; }
p { margin: 0; }

:where(button, input, select, summary, [tabindex]):focus-visible {
  outline: 2px solid var(--accent);
  outline-offset: 2px;
  border-radius: 6px;
}
.sr { position: absolute; width: 1px; height: 1px; padding: 0; margin: -1px;
      overflow: hidden; clip: rect(0 0 0 0); white-space: nowrap; border: 0; }

/* ---------- header ---------- */
header { margin-bottom: 16px; }
.eyebrow {
  font-family: var(--font-mono); font-size: 11px; letter-spacing: .12em;
  text-transform: uppercase; color: var(--muted); display: flex;
  align-items: center; gap: 8px; flex-wrap: wrap;
}
.eyebrow .dot { width: 5px; height: 5px; border-radius: 50%; background: var(--accent); flex: none; }
h1 { font-size: clamp(21px, 4.4vw, 29px); margin: 6px 0 6px; }
.lede { color: var(--muted); font-size: 15px; max-width: 64ch; text-wrap: pretty; }
.warnbar {
  margin-top: 12px; display: flex; gap: 8px; align-items: baseline; flex-wrap: wrap;
  border: 1px solid var(--line); border-left: 3px solid var(--accent);
  background: var(--accent-soft); border-radius: 4px; padding: 8px 11px;
}
.warnbar b { font-family: var(--font-mono); font-size: 11px; letter-spacing: .1em;
  text-transform: uppercase; color: var(--accent); }
.warnbar span { font-size: 13px; color: var(--ink); }

/* ---------- layout ---------- */
.grid { display: grid; grid-template-columns: minmax(0, 1.6fr) minmax(0, 1fr); gap: 14px; align-items: start; }
@media (max-width: 900px) {
  /* one flat column: chart, numbers, how it reads, then the data you edit.
     the order has to track source order or tabbing jumps around the page */
  .grid { grid-template-columns: minmax(0, 1fr); }
  .grid > .col { display: contents; }
  #cChart { order: 1; } #cReadout { order: 2; } #cMethod { order: 3; }
  #cData { order: 4; } #cSets { order: 5; }
}
@media (max-width: 430px) { .tiles > .tile:last-child { grid-column: 1 / -1; } }
.col { display: grid; gap: 14px; min-width: 0; }

.card {
  background: var(--surface); border: 1px solid var(--line);
  border-radius: var(--radius); padding: 13px; min-width: 0;
}
.card > h2 {
  font-family: var(--font-mono); font-size: 11px; letter-spacing: .12em;
  text-transform: uppercase; color: var(--muted); font-weight: 600;
  display: flex; justify-content: space-between; align-items: center; gap: 8px;
  margin-bottom: 10px;
}

/* ---------- chart ---------- */
.chartbox { position: relative; }
canvas { display: block; width: 100%; height: auto; touch-action: pan-y; }
.hoverline {
  font-family: var(--font-mono); font-size: 11.5px; color: var(--muted);
  margin-top: 6px; min-height: 17px; white-space: nowrap; overflow: hidden;
  text-overflow: ellipsis;
}
.hoverline b { color: var(--ink); font-weight: 600; }
@media (max-width: 620px) {
  /* let the point readout wrap — clipping it drops the residual, the useful half */
  .hoverline { white-space: normal; overflow: visible; min-height: 34px; }
}
.legend { display: flex; gap: 13px; flex-wrap: wrap; font-family: var(--font-mono);
  font-size: 11px; color: var(--muted); margin-top: 2px; }
.legend i { display: inline-block; width: 14px; height: 3px; border-radius: 2px;
  vertical-align: middle; margin-right: 5px; background: var(--accent); }
.legend .pt { width: 8px; height: 8px; border-radius: 50%; background: var(--ink); }
.legend .band { background: var(--accent); opacity: .3; height: 9px; }

/* ---------- readouts ---------- */
.tiles { display: grid; grid-template-columns: repeat(auto-fit, minmax(104px, 1fr)); gap: 1px;
  background: var(--line); border: 1px solid var(--line); border-radius: 8px; overflow: hidden; }
.tile { background: var(--surface); padding: 9px 10px; min-width: 0; }
.tile .k { font-family: var(--font-mono); font-size: 10px; letter-spacing: .1em;
  text-transform: uppercase; color: var(--muted); }
.tile .v { font-family: var(--font-mono); font-size: 19px; font-weight: 600;
  line-height: 1.25; margin-top: 2px; overflow-wrap: anywhere; }
.tile .v small { font-size: 11px; font-weight: 500; color: var(--muted); }
.tile .s { font-family: var(--font-mono); font-size: 10.5px; letter-spacing: -.01em;
  color: var(--muted); margin-top: 1px; }
.tile.hero .v { color: var(--accent); font-size: 23px; }

.chips { display: flex; flex-wrap: wrap; gap: 6px; margin-top: 10px;
  font-family: var(--font-mono); font-size: 11px; }
.chip { border: 1px solid var(--line); border-radius: 999px; padding: 2px 9px; color: var(--muted); }
.chip.ok { color: var(--run); border-color: var(--run); background: var(--run-soft); }
.chip.bad { color: var(--danger); border-color: var(--danger); }
.chip b { font-weight: 600; color: var(--ink); }
.chip.ok b { color: var(--run); }

/* ---------- table ---------- */
table { width: 100%; border-collapse: collapse; table-layout: fixed; }
caption { text-align: left; }
th { font-family: var(--font-mono); font-size: 10px; letter-spacing: .08em;
  text-transform: uppercase; color: var(--muted); font-weight: 600;
  text-align: left; padding: 0 4px 6px; }
th.n { width: 24px; padding-left: 0; }
th.r { width: 52px; text-align: right; }
th.x { width: 34px; }
td { padding: 2px 4px; }
td.n { font-family: var(--font-mono); font-size: 11px; color: var(--muted); padding-left: 0; }
td.r { font-family: var(--font-mono); font-size: 11.5px; text-align: right; color: var(--muted); }
tr.hi td.n { color: var(--accent); font-weight: 700; }
tr.hi input { border-color: var(--accent); }
input[type=number] {
  width: 100%; min-width: 0; font-family: var(--font-mono); font-size: 13px;
  padding: 5px 6px; border: 1px solid var(--line); border-radius: 6px;
  background: var(--bg); color: var(--ink); -moz-appearance: textfield;
}
input[type=number]::-webkit-outer-spin-button,
input[type=number]::-webkit-inner-spin-button { -webkit-appearance: none; margin: 0; }
input[type=number]:hover { border-color: var(--muted); }
input.invalid { border-color: var(--danger); color: var(--danger); }
.xbtn {
  width: 30px; height: 30px; display: grid; place-items: center; padding: 0;
  border: 1px solid transparent; border-radius: 6px; background: none;
  color: var(--muted); font-family: var(--font-mono); font-size: 14px; cursor: pointer;
}
.xbtn:hover { border-color: var(--danger); color: var(--danger); }

/* ---------- controls ---------- */
.row { display: flex; flex-wrap: wrap; gap: 7px; align-items: center; }
.row + .row { margin-top: 9px; }
button.b {
  font-family: var(--font-mono); font-size: 12px; padding: 6px 11px; cursor: pointer;
  border: 1px solid var(--line); border-radius: 7px; background: var(--surface);
  color: var(--ink); transition: border-color .12s, background .12s;
}
button.b:hover:not(:disabled) { border-color: var(--accent); }
button.b[aria-pressed="true"] { border-color: var(--accent); background: var(--accent-soft); color: var(--accent); }
button.b:disabled { opacity: .45; cursor: not-allowed; }
button.b.primary { border-color: var(--accent); color: var(--accent); background: var(--accent-soft); }
label.ck { display: flex; align-items: center; gap: 7px; font-size: 13px; cursor: pointer; }
label.ck input { accent-color: var(--accent); width: 15px; height: 15px; }
select {
  font-family: var(--font-mono); font-size: 12px; padding: 5px 7px;
  border: 1px solid var(--line); border-radius: 7px; background: var(--surface); color: var(--ink);
}
select:disabled { opacity: .45; cursor: not-allowed; }
label.ck:has(input:disabled) { opacity: .45; cursor: not-allowed; }
label.ck input:disabled { cursor: not-allowed; }
.lbl { font-family: var(--font-mono); font-size: 11px; letter-spacing: .06em;
  text-transform: uppercase; color: var(--muted); }
.hint { font-size: 12.5px; color: var(--muted); margin-top: 9px; }

/* ---------- loading / error ---------- */
.boot { border: 1px solid var(--line); border-radius: var(--radius); padding: 13px;
  background: var(--surface); }
.boot .stage { font-family: var(--font-mono); font-size: 12px; display: flex;
  justify-content: space-between; gap: 10px; }
.boot .stage .pct { color: var(--muted); }
.bar { height: 4px; border-radius: 999px; background: var(--line); margin-top: 8px; overflow: hidden; }
.bar > i { display: block; height: 100%; width: 0%; background: var(--accent);
  transition: width .3s ease; }
.bar.done > i { background: var(--run); }
.boot .sub { font-size: 12.5px; color: var(--muted); margin-top: 7px; }
.fail { border-color: var(--danger); }
.fail .stage { color: var(--danger); }

/* ---------- method ---------- */
details.method { border: 1px solid var(--line); border-radius: var(--radius);
  background: var(--surface); padding: 11px 13px; }
details.method summary { font-family: var(--font-mono); font-size: 11px;
  letter-spacing: .12em; text-transform: uppercase; color: var(--muted);
  cursor: pointer; font-weight: 600; }
details.method summary::marker { color: var(--accent); }
.method .body { margin-top: 10px; font-size: 13.5px; color: var(--muted); display: grid; gap: 9px; }
.method .body b { color: var(--ink); font-weight: 600; }
.eq { font-family: var(--font-mono); font-size: 13px; color: var(--ink);
  background: var(--bg); border: 1px solid var(--line); border-radius: 7px;
  padding: 9px 11px; overflow-x: auto; white-space: pre; }

footer { margin-top: 22px; padding-top: 12px; border-top: 1px solid var(--line);
  font-family: var(--font-mono); font-size: 11px; color: var(--muted);
  display: flex; justify-content: space-between; gap: 10px; flex-wrap: wrap; }

@media (prefers-reduced-motion: reduce) {
  * { transition-duration: .01ms !important; animation-duration: .01ms !important; }
}
</style>
</head>
<body>
<div class="wrap">

<header>
  <div class="eyebrow"><span class="dot" aria-hidden="true"></span>pharma · pyodide + numpy · no libraries</div>
  <h1>Dose-Response Curve Fitter</h1>
  <p class="lede">Type in concentrations and % response; Python fits a 4-parameter Hill curve
  by hand-rolled Nelder-Mead and bootstraps the IC<sub>50</sub> confidence interval — all in your browser.</p>
  <div class="warnbar">
    <b>teaching tool, not a lab tool</b>
    <span>It shows how curve fitting and its uncertainty behave. Do not use it for regulated or reportable results.</span>
  </div>
</header>

<div id="boot" class="boot" role="status" aria-live="polite">
  <div class="stage"><span id="bootStage">starting python runtime…</span><span class="pct" id="bootPct">0%</span></div>
  <div class="bar" id="bootBar"><i></i></div>
  <div class="sub" id="bootSub">CPython + numpy compiled to WebAssembly. First load pulls the runtime once, then it is cached.</div>
</div>

<div class="grid" id="app">
  <div class="col">

    <section class="card" id="cChart">
      <h2><span>fitted curve</span><span id="chartTag">8 points · log scale</span></h2>
      <div class="chartbox">
        <canvas id="plot" role="img" aria-label="Dose-response scatter plot with fitted curve"></canvas>
      </div>
      <div class="hoverline" id="hover">hover the chart, or focus a row below, to inspect a point</div>
      <div class="legend" aria-hidden="true">
        <span><i class="pt"></i>measured</span>
        <span><i></i>Hill fit</span>
        <span><i class="band"></i>bootstrap 95% band</span>
        <span><i style="background:var(--muted)"></i>IC50</span>
      </div>
    </section>

    <section class="card" id="cReadout">
      <h2><span>readout</span><span id="fitTime">—</span></h2>
      <div class="tiles">
        <div class="tile hero"><div class="k">IC50</div><div class="v" id="oIc50">—</div><div class="s" id="oIc50ci">95% CI —</div></div>
        <div class="tile"><div class="k">Hill slope</div><div class="v" id="oHill">—</div><div class="s" id="oHillci">95% CI —</div></div>
        <div class="tile"><div class="k">Top</div><div class="v" id="oTop">—</div><div class="s">% response</div></div>
        <div class="tile"><div class="k">Bottom</div><div class="v" id="oBot">—</div><div class="s">% response</div></div>
        <div class="tile"><div class="k">R²</div><div class="v" id="oR2">—</div><div class="s" id="oRmse">RMSE —</div></div>
      </div>
      <div class="chips" id="chips"></div>
      <p class="hint" id="note"></p>
    </section>

    <details class="method" id="cMethod">
      <summary>how it works</summary>
      <div class="body">
        <div>The model is the <b>4-parameter Hill (logistic) equation</b> — the standard shape for a
        competitive dose-response:</div>
        <div class="eq">response(x) = bottom + (top - bottom) / (1 + (x / IC50) ^ hill)</div>
        <div><b>Fitting.</b> Least squares on the four parameters, minimised by a Nelder-Mead simplex
        written from scratch in numpy — no scipy, no solver library. IC<sub>50</sub> is fitted in log10 space
        (that is the scale the data lives on) and the model is evaluated as
        <span style="font-family:var(--font-mono)">1/(1+exp(hill·ln10·(log10 x − log10 IC50)))</span>
        so it never overflows. Three starting guesses (hill = 0.5, 1, 2) are tried and the best kept.</div>
        <div><b>Uncertainty.</b> Residual bootstrap: the residuals from the best fit are resampled with
        replacement, added back onto the fitted curve, and the whole fit is re-run — a few hundred times.
        The 2.5th and 97.5th percentiles of those refits give the CI and the shaded band. Residual
        resampling (rather than resampling points) is used because an 8-point dilution series is a
        designed set — dropping a point at random would destroy the design.</div>
        <div><b>Why it is fast.</b> Running hundreds of simplexes one after another in Python takes seconds.
        Instead every resample gets its own simplex and they all march in lockstep: one reflection, one
        expansion and two contractions for the entire batch, as four numpy calls per iteration on a
        <span style="font-family:var(--font-mono)">B × 5 × 4</span> array. Same algorithm, one pass, and the
        whole bootstrap lands in a fraction of a second.</div>
        <div><b>Why it matters.</b> A single IC<sub>50</sub> number hides how much the data actually pins it down.
        Push the curve so its bottom plateau is off the edge of the tested range and watch the CI blow up
        while R² stays comfortably high.</div>
      </div>
    </details>

  </div>

  <div class="col">

    <section class="card" id="cData">
      <h2><span>data</span><span id="dataTag">nM · % response</span></h2>
      <table>
        <caption class="sr">Editable assay data: concentration in nanomolar and percent response, with fit residual per point.</caption>
        <thead>
          <tr>
            <th class="n" scope="col"><span class="sr">point</span>#</th>
            <th scope="col">conc nM</th>
            <th scope="col">% resp</th>
            <th class="r" scope="col">resid</th>
            <th class="x" scope="col"><span class="sr">remove</span></th>
          </tr>
        </thead>
        <tbody id="tbody"></tbody>
      </table>
      <div class="row" style="margin-top:10px">
        <button class="b" id="addPt" type="button">+ add point</button>
        <button class="b primary" id="refit" type="button">refit ▸</button>
      </div>
      <p class="hint" id="dataNote">Edit any cell — the curve refits as you type.</p>
    </section>

    <section class="card" id="cSets">
      <h2><span>datasets</span></h2>
      <div class="row" role="group" aria-label="Example datasets" id="presets"></div>
      <div class="row">
        <span class="lbl">bootstrap</span>
        <select id="bootN" aria-label="Number of bootstrap resamples">
          <option value="100">100 resamples</option>
          <option value="200" selected>200 resamples</option>
          <option value="500">500 resamples</option>
          <option value="1000">1000 resamples</option>
        </select>
      </div>
      <div class="row">
        <label class="ck"><input type="checkbox" id="fixBottom"> constrain bottom = 0%</label>
      </div>
      <p class="hint">Constraining the bottom plateau is what you do when the assay never reaches full
      inhibition — it trades a free parameter for a much tighter IC<sub>50</sub>. Try it on <b>compound C</b>.</p>
    </section>

  </div>
</div>

<footer>
  <span>demo · sloprun.dev</span>
  <span id="engine">python: —</span>
</footer>

</div>

<script type="text/python" id="pysrc">
import json, math, time
import numpy as np

LN10 = 2.302585092994046
STATE = {}


def _clean(o):
    """NaN / Infinity are not valid JSON — emit null rather than a payload
    JSON.parse will reject (all-identical responses give an undefined R2)."""
    if isinstance(o, float):
        return o if math.isfinite(o) else None
    if isinstance(o, list):
        return [_clean(v) for v in o]
    if isinstance(o, dict):
        return dict((k, _clean(v)) for k, v in o.items())
    return o


def jd(d):
    return json.dumps(_clean(d), allow_nan=False)


def model(lx, p):
    """4-parameter Hill, evaluated in log space so it never overflows."""
    top, bot, lic, h = p[0], p[1], p[2], p[3]
    t = np.clip(h * (lx - lic) * LN10, -500.0, 500.0)
    return bot + (top - bot) / (1.0 + np.exp(t))


def expand(p, fix_bottom):
    if fix_bottom:
        return np.array([p[0], 0.0, p[1], p[2]])
    return p


def soft(v, lo, hi, w):
    if v < lo:
        return w * (lo - v) ** 2
    if v > hi:
        return w * (v - hi) ** 2
    return 0.0


def make_obj(lx, y, fix_bottom, lo, hi):
    def obj(p):
        q = expand(p, fix_bottom)
        pen = soft(q[3], 0.05, 12.0, 1e4)
        pen += soft(q[2], lo - 2.0, hi + 2.0, 1e4)
        pen += soft(q[0], -60.0, 220.0, 1e2)
        pen += soft(q[1], -60.0, 220.0, 1e2)
        r = y - model(lx, q)
        return float(np.dot(r, r)) + pen
    return obj


def nelder_mead(f, x0, steps, maxiter=700, ftol=1e-12, xtol=1e-10):
    x0 = np.asarray(x0, dtype=float)
    n = x0.size
    sim = np.tile(x0, (n + 1, 1))
    for i in range(n):
        sim[i + 1, i] = x0[i] + steps[i]
    fv = np.array([f(s) for s in sim])
    nev = n + 1
    it = 0
    conv = False
    while it < maxiter:
        o = np.argsort(fv)
        sim = sim[o]
        fv = fv[o]
        if (np.max(np.abs(fv[1:] - fv[0])) <= ftol * (abs(fv[0]) + ftol)
                and np.max(np.abs(sim[1:] - sim[0])) <= xtol * (np.max(np.abs(sim[0])) + xtol)):
            conv = True
            break
        it += 1
        cen = np.mean(sim[:-1], axis=0)
        d = cen - sim[-1]
        xr = cen + d
        fr = f(xr)
        nev += 1
        if fr < fv[0]:
            xe = cen + 2.0 * d
            fe = f(xe)
            nev += 1
            if fe < fr:
                sim[-1], fv[-1] = xe, fe
            else:
                sim[-1], fv[-1] = xr, fr
        elif fr < fv[-2]:
            sim[-1], fv[-1] = xr, fr
        else:
            if fr < fv[-1]:
                xc = cen + 0.5 * d
                fc = f(xc)
                nev += 1
                ok = fc <= fr
            else:
                xc = cen - 0.5 * d
                fc = f(xc)
                nev += 1
                ok = fc < fv[-1]
            if ok:
                sim[-1], fv[-1] = xc, fc
            else:
                sim[1:] = sim[0] + 0.5 * (sim[1:] - sim[0])
                fv[1:] = np.array([f(s) for s in sim[1:]])
                nev += n
    o = np.argsort(fv)
    return sim[o][0], float(fv[o][0]), it, nev, conv


def guess(lx, y):
    o = np.argsort(lx)
    ys = y[o]
    ls = lx[o]
    k = max(1, len(ys) // 4)
    top = float(np.mean(ys[:k]))
    bot = float(np.mean(ys[-k:]))
    mid = 0.5 * (top + bot)
    lic = float(np.median(ls))
    for i in range(len(ys) - 1):
        a, b = ys[i], ys[i + 1]
        if (a - mid) * (b - mid) <= 0 and a != b:
            lic = float(ls[i] + (a - mid) / (a - b) * (ls[i + 1] - ls[i]))
            break
    return top, bot, lic


def core_fit(lx, y, fix_bottom, start=None):
    lo, hi = float(np.min(lx)), float(np.max(lx))
    obj = make_obj(lx, y, fix_bottom, lo, hi)
    best = None
    if start is not None:
        s = np.array(start, dtype=float)
        steps = np.maximum(np.abs(s) * 0.06, 0.06)
        best = nelder_mead(obj, s, steps, maxiter=400)
    else:
        top, bot, lic = guess(lx, y)
        for h0 in (0.5, 1.0, 2.0):
            if fix_bottom:
                p0 = np.array([top, lic, h0])
                steps = np.array([max(5.0, abs(top) * 0.1), 0.4, 0.35])
            else:
                p0 = np.array([top, bot, lic, h0])
                steps = np.array([max(5.0, abs(top) * 0.1), max(5.0, abs(bot) * 0.1 + 3.0), 0.4, 0.35])
            r = nelder_mead(obj, p0, steps)
            if best is None or r[1] < best[1]:
                best = r
    return best


def fit(payload):
    t0 = time.perf_counter()
    d = json.loads(payload)
    x = np.array(d["x"], dtype=float)
    y = np.array(d["y"], dtype=float)
    fix_bottom = bool(d["fix_bottom"])
    lx = np.log10(x)
    p, sse, it, nev, conv = core_fit(lx, y, fix_bottom)
    q = expand(p, fix_bottom)
    yhat = model(lx, q)
    resid = y - yhat
    sse = float(np.dot(resid, resid))
    sst = float(np.dot(y - np.mean(y), y - np.mean(y)))
    r2 = 1.0 - sse / sst if sst > 0 else float("nan")
    k = 3 if fix_bottom else 4
    dof = max(1, len(y) - k)
    rmse = float(np.sqrt(sse / dof))
    g0, g1 = float(np.min(lx)) - 0.6, float(np.max(lx)) + 0.6
    grid = np.linspace(g0, g1, 180)
    curve = model(grid, q)
    STATE.clear()
    STATE.update(dict(lx=lx, y=y, yhat=yhat, resid=resid - float(np.mean(resid)),
                      fix_bottom=fix_bottom, p=p, grid=grid))
    return jd(dict(
        ok=True, converged=bool(conv),
        top=float(q[0]), bottom=float(q[1]), ic50=float(10.0 ** q[2]),
        logic50=float(q[2]), hill=float(q[3]),
        sse=float(sse), r2=float(r2), rmse=rmse, dof=dof,
        iters=int(it), nev=int(nev), ms=round((time.perf_counter() - t0) * 1000.0, 1),
        resid=[round(float(v), 3) for v in resid],
        grid=[round(float(v), 5) for v in grid],
        curve=[round(float(v), 4) for v in curve],
    ))


def expand_batch(P, fix_bottom):
    """(K, n) parameter rows -> (K, 4) as [top, bottom, log10 IC50, hill]."""
    if fix_bottom:
        z = np.zeros((P.shape[0], 1))
        return np.hstack([P[:, 0:1], z, P[:, 1:2], P[:, 2:3]])
    return P


def model_batch(lx, Q):
    """One curve per parameter row: (K, 4) x (m,) -> (K, m)."""
    t = np.clip(Q[:, 3:4] * (lx[None, :] - Q[:, 2:3]) * LN10, -500.0, 500.0)
    return Q[:, 1:2] + (Q[:, 0:1] - Q[:, 1:2]) / (1.0 + np.exp(t))


def soft_b(v, lo, hi, w):
    return w * (np.maximum(lo - v, 0.0) ** 2 + np.maximum(v - hi, 0.0) ** 2)


def batch_sse(P, ridx, lx, Y, fix_bottom, lo, hi):
    """Objective for K candidate parameter rows, each scored against dataset Y[ridx]."""
    q = expand_batch(P, fix_bottom)
    r = Y[ridx] - model_batch(lx, q)
    pen = (soft_b(q[:, 3], 0.05, 12.0, 1e4)
           + soft_b(q[:, 2], lo - 2.0, hi + 2.0, 1e4)
           + soft_b(q[:, 0], -60.0, 220.0, 1e2)
           + soft_b(q[:, 1], -60.0, 220.0, 1e2))
    return np.sum(r * r, axis=1) + pen


def nelder_mead_batch(P0, steps, lx, Y, fix_bottom, lo, hi, maxiter=600,
                      ftol=1e-10, xtol=1e-8):
    """B independent simplexes marched in lockstep — every reflection, expansion
    and contraction for the whole bootstrap is four numpy calls per iteration."""
    B, n = P0.shape
    m = n + 1
    sim = np.repeat(P0[:, None, :], m, axis=1)
    for i in range(n):
        sim[:, i + 1, i] += steps[i]
    rows = np.arange(B)
    fv = batch_sse(sim.reshape(B * m, n), np.repeat(rows, m),
                   lx, Y, fix_bottom, lo, hi).reshape(B, m)
    done = np.zeros(B, dtype=bool)
    it = 0
    while it < maxiter:
        o = np.argsort(fv, axis=1)
        sim = np.take_along_axis(sim, o[:, :, None], axis=1)
        fv = np.take_along_axis(fv, o, axis=1)
        sf = np.max(np.abs(fv[:, 1:] - fv[:, 0:1]), axis=1)
        sx = np.max(np.abs(sim[:, 1:, :] - sim[:, 0:1, :]), axis=(1, 2))
        done = ((sf <= ftol * (np.abs(fv[:, 0]) + ftol))
                & (sx <= xtol * (np.max(np.abs(sim[:, 0, :]), axis=1) + xtol)))
        if bool(np.all(done)):
            break
        it += 1
        cen = np.mean(sim[:, :-1, :], axis=1)
        d = cen - sim[:, -1, :]
        xr = cen + d
        xe = cen + 2.0 * d
        xoc = cen + 0.5 * d
        xic = cen - 0.5 * d
        fr = batch_sse(xr, rows, lx, Y, fix_bottom, lo, hi)
        fe = batch_sse(xe, rows, lx, Y, fix_bottom, lo, hi)
        foc = batch_sse(xoc, rows, lx, Y, fix_bottom, lo, hi)
        fic = batch_sse(xic, rows, lx, Y, fix_bottom, lo, hi)

        best, second, worst = fv[:, 0], fv[:, -2], fv[:, -1]
        m_exp = fr < best
        m_ref = (~m_exp) & (fr < second)
        rest = (~m_exp) & (~m_ref)
        m_oc = rest & (fr < worst) & (foc <= fr)
        m_ic = rest & (fr >= worst) & (fic < worst)
        m_shrink = ~(m_exp | m_ref | m_oc | m_ic)

        cx, cf = xr.copy(), fr.copy()
        ue = m_exp & (fe < fr)
        cx[ue], cf[ue] = xe[ue], fe[ue]
        cx[m_oc], cf[m_oc] = xoc[m_oc], foc[m_oc]
        cx[m_ic], cf[m_ic] = xic[m_ic], fic[m_ic]

        keep = (~m_shrink) & (~done)
        sim[keep, -1, :] = cx[keep]
        fv[keep, -1] = cf[keep]

        sh = m_shrink & (~done)
        k = int(np.count_nonzero(sh))
        if k:
            base = sim[sh, 0:1, :]
            sim[sh, 1:, :] = base + 0.5 * (sim[sh, 1:, :] - base)
            pts = sim[sh, 1:, :].reshape(k * n, n)
            fv[sh, 1:] = batch_sse(pts, np.repeat(rows[sh], n),
                                   lx, Y, fix_bottom, lo, hi).reshape(k, n)
    o = np.argsort(fv, axis=1)
    bestx = np.take_along_axis(sim, o[:, :, None], axis=1)[:, 0, :]
    return bestx, int(np.count_nonzero(~done)), it


def boot_start(n, seed):
    n = int(n)
    STATE["rng"] = np.random.default_rng(int(seed))
    STATE["n"] = n
    STATE["done"] = 0
    STATE["lic"] = np.zeros(n)
    STATE["hill"] = np.zeros(n)
    STATE["bad"] = 0
    STATE["rounds"] = 0
    STATE["curves"] = np.zeros((n, STATE["grid"].size))
    STATE["t0"] = time.perf_counter()
    return n


def boot_run(k):
    """Resample residuals for the next chunk and refit that whole chunk at once."""
    lx, yhat, res = STATE["lx"], STATE["yhat"], STATE["resid"]
    fix_bottom, p0, grid = STATE["fix_bottom"], STATE["p"], STATE["grid"]
    rng, m = STATE["rng"], yhat.size
    i = STATE["done"]
    end = min(STATE["n"], i + int(k))
    B = end - i
    if B <= 0:
        return i
    Y = yhat[None, :] + res[rng.integers(0, m, (B, m))]
    p0 = np.asarray(p0, dtype=float)
    P0 = np.tile(p0, (B, 1))
    steps = np.maximum(np.abs(p0) * 0.06, 0.06)
    lo, hi = float(np.min(lx)), float(np.max(lx))
    P, bad, it = nelder_mead_batch(P0, steps, lx, Y, fix_bottom, lo, hi)
    Q = expand_batch(P, fix_bottom)
    STATE["lic"][i:end] = Q[:, 2]
    STATE["hill"][i:end] = Q[:, 3]
    STATE["curves"][i:end] = model_batch(grid, Q)
    STATE["bad"] += bad
    STATE["rounds"] = max(STATE["rounds"], it)
    STATE["done"] = end
    return end


def boot_finish():
    lic, hill, curves = STATE["lic"], STATE["hill"], STATE["curves"]
    lo, hi = np.percentile(lic, [2.5, 97.5])
    hlo, hhi = np.percentile(hill, [2.5, 97.5])
    blo = np.percentile(curves, 2.5, axis=0)
    bhi = np.percentile(curves, 97.5, axis=0)
    return jd(dict(
        n=int(STATE["n"]),
        ic50_lo=float(10.0 ** lo), ic50_hi=float(10.0 ** hi),
        hill_lo=float(hlo), hill_hi=float(hhi),
        nonconv=int(STATE["bad"]), rounds=int(STATE["rounds"]),
        ms=round((time.perf_counter() - STATE["t0"]) * 1000.0, 1),
        band_lo=[round(float(v), 4) for v in blo],
        band_hi=[round(float(v), 4) for v in bhi],
    ))
</script>

<script>
(function () {
  'use strict';

  /* ---------------- data ---------------- */
  var CONC = [0.1, 0.3, 1, 3, 10, 30, 100, 300];
  var PRESETS = [
    { id: 'A', label: 'compound A · potent',
      note: 'A clean 8-point half-log series. The curve is fully bracketed: both plateaus are measured, so the IC50 interval is tight.',
      x: CONC, y: [101.2, 97.4, 96.8, 82.1, 58.9, 25.4, 11.7, 4.2] },
    { id: 'B', label: 'compound B · steep',
      note: 'Hill slope near 3 — cooperative, switch-like. Steep curves pin the IC50 down hard but leave the plateaus poorly sampled.',
      x: CONC, y: [99.6, 101.1, 98.4, 100.7, 96.2, 73.5, 6.8, 1.5] },
    { id: 'C', label: 'compound C · unbracketed',
      note: 'Weak and noisy: the bottom plateau is off the right edge of the tested range. R² still looks fine — the CI does not. Try constraining bottom = 0%.',
      x: CONC, y: [92.0, 105.3, 91.4, 102.9, 83.7, 86.1, 55.0, 42.9] }
  ];

  var DATA = PRESETS[0].x.map(function (v, i) { return { x: v, y: PRESETS[0].y[i] }; });
  var activePreset = 'A';
  var FIT = null, BOOT = null, hoverIdx = -1, curveT = 1, fixBottom = false, bootN = 200;
  var py = null, pyfn = {}, busy = false, runToken = 0;

  var $ = function (id) { return document.getElementById(id); };
  var reduceMotion = matchMedia('(prefers-reduced-motion: reduce)').matches;

  /* ---------------- formatting ---------------- */
  function fin(v) { return typeof v === 'number' && isFinite(v); }
  function sig(v, n) {
    if (!fin(v)) return '—';
    var a = Math.abs(v);
    if (a === 0) return '0';
    if (a >= 1e5 || a < 1e-3) return v.toExponential(Math.max(0, n - 1));
    var d = Math.max(0, n - 1 - Math.floor(Math.log10(a)));
    return v.toFixed(Math.min(d, 6));
  }
  function num(v, d) { return fin(v) ? v.toFixed(d) : '—'; }

  /* ---------------- table ---------------- */
  function renderTable() {
    var tb = $('tbody');
    tb.textContent = '';
    DATA.forEach(function (pt, i) {
      var tr = document.createElement('tr');
      tr.dataset.i = i;

      var tdn = document.createElement('td');
      tdn.className = 'n'; tdn.textContent = String(i + 1);
      tr.appendChild(tdn);

      tr.appendChild(cell(i, 'x', pt.x, 'concentration in nanomolar for point ' + (i + 1)));
      tr.appendChild(cell(i, 'y', pt.y, 'percent response for point ' + (i + 1)));

      var tdr = document.createElement('td');
      tdr.className = 'r'; tdr.dataset.resid = i; tdr.textContent = '—';
      tr.appendChild(tdr);

      var tdx = document.createElement('td');
      var b = document.createElement('button');
      b.type = 'button'; b.className = 'xbtn'; b.textContent = '✕';
      b.setAttribute('aria-label', 'remove point ' + (i + 1));
      b.addEventListener('click', function () {
        if (DATA.length <= 4) { flashNote('need at least 4 points to fit 4 parameters'); return; }
        DATA.splice(i, 1); activePreset = null; renderTable(); markPresets(); scheduleFit(0);
      });
      tdx.appendChild(b);
      tr.appendChild(tdx);
      tb.appendChild(tr);
    });
    $('chartTag').textContent = DATA.length + ' points · log scale';
  }

  function cell(i, key, val, label) {
    var td = document.createElement('td');
    var inp = document.createElement('input');
    inp.type = 'number'; inp.inputMode = 'decimal'; inp.step = 'any';
    inp.value = String(val);
    inp.setAttribute('aria-label', label);
    inp.addEventListener('input', function () {
      var v = parseFloat(inp.value);
      var bad = !isFinite(v) || (key === 'x' && v <= 0);
      inp.classList.toggle('invalid', bad);
      /* an unusable cell drops its point from the fit rather than quietly fitting a stale value */
      DATA[i][key] = bad ? NaN : v;
      activePreset = null; markPresets();
      var td = $('tbody').querySelector('[data-resid="' + i + '"]');
      if (bad && td) td.textContent = '·';
      scheduleFit(bad ? 700 : 380);
      syncHover();
    });
    inp.addEventListener('focus', function () { hoverIdx = i; syncHover(); });
    inp.addEventListener('blur', function () { if (hoverIdx === i) { hoverIdx = -1; syncHover(); } });
    td.appendChild(inp);
    return td;
  }

  function validPoints() {
    var out = [];
    DATA.forEach(function (p, i) {
      if (isFinite(p.x) && isFinite(p.y) && p.x > 0) out.push({ x: p.x, y: p.y, i: i });
    });
    return out;
  }

  var noteTimer = null;
  var noteIdle = 'Edit any cell — the curve refits as you type.';
  function flashNote(msg) {
    var n = $('dataNote');
    clearTimeout(noteTimer);
    n.textContent = msg;
    n.style.color = 'var(--danger)';
    noteTimer = setTimeout(function () {
      n.textContent = noteIdle;
      n.style.color = '';
    }, 2600);
  }

  /* ---------------- presets ---------------- */
  function buildPresets() {
    var box = $('presets');
    PRESETS.forEach(function (p) {
      var b = document.createElement('button');
      b.type = 'button'; b.className = 'b'; b.textContent = p.label;
      b.dataset.p = p.id;
      b.addEventListener('click', function () {
        DATA = p.x.map(function (v, i) { return { x: v, y: p.y[i] }; });
        activePreset = p.id;
        renderTable(); markPresets(); $('note').textContent = p.note; scheduleFit(0);
      });
      box.appendChild(b);
    });
    markPresets();
  }
  function markPresets() {
    Array.prototype.forEach.call($('presets').children, function (b) {
      b.setAttribute('aria-pressed', String(b.dataset.p === activePreset));
    });
  }

  /* ---------------- canvas ---------------- */
  var cv = $('plot'), ctx = cv.getContext('2d'), PLOT = null;

  function css(name) {
    return getComputedStyle(document.documentElement).getPropertyValue(name).trim();
  }
  function rgba(hex, a) {
    var h = hex.replace('#', '');
    if (h.length === 3) h = h[0] + h[0] + h[1] + h[1] + h[2] + h[2];
    var n = parseInt(h, 16);
    return 'rgba(' + ((n >> 16) & 255) + ',' + ((n >> 8) & 255) + ',' + (n & 255) + ',' + a + ')';
  }

  /* y labels have a 37px gutter — a typo like 1e9 in a response cell must not
     render as a clipped row of zeros sitting on top of the axis title */
  function trimz(s) {
    return s.indexOf('.') < 0 ? s : s.replace(/0+$/, '').replace(/\.$/, '');
  }
  function axisNum(v) {
    var a = Math.abs(v);
    if (a >= 1e9) return v.toExponential(1);
    if (a >= 1e6) return trimz((v / 1e6).toPrecision(3)) + 'M';
    if (a >= 1e4) return trimz((v / 1e3).toPrecision(3)) + 'k';
    return trimz(String(Math.round(v * 1e4) / 1e4));
  }

  function niceStep(range, target) {
    var raw = range / Math.max(1, target);
    if (!isFinite(raw) || raw <= 0) return 1;
    var mag = Math.pow(10, Math.floor(Math.log10(raw)));
    var k = raw / mag;
    return (k <= 1 ? 1 : k <= 2 ? 2 : k <= 5 ? 5 : 10) * mag;
  }

  function draw() {
    var box = cv.parentElement.getBoundingClientRect();
    var w = Math.max(240, Math.round(box.width));
    var h = Math.round(Math.min(390, Math.max(230, w * 0.62)));
    var dpr = Math.min(2.5, window.devicePixelRatio || 1);
    if (cv.width !== Math.round(w * dpr) || cv.height !== Math.round(h * dpr)) {
      cv.width = Math.round(w * dpr); cv.height = Math.round(h * dpr);
    }
    cv.style.height = h + 'px';
    ctx.setTransform(dpr, 0, 0, dpr, 0, 0);
    ctx.clearRect(0, 0, w, h);

    var ink = css('--ink'), muted = css('--muted'), line = css('--line'),
        accent = css('--accent'), surface = css('--surface');
    ctx.fillStyle = surface; ctx.fillRect(0, 0, w, h);

    var pts = validPoints();
    var L = 44, R = 12, T = 12, B = 34;
    var pw = Math.max(20, w - L - R), ph = Math.max(20, h - T - B);

    var lxs = pts.map(function (p) { return Math.log10(p.x); });
    /* an all-identical column makes R2 undefined, and python sends null rather than
       NaN — every value off the fit is checked before it can poison a coordinate */
    var fitOk = !!FIT && fin(FIT.top) && fin(FIT.bottom) && fin(FIT.logic50);
    var gridOk = !!FIT && fin(FIT.grid[0]) && fin(FIT.grid[FIT.grid.length - 1]);
    var lo, hi;
    if (gridOk) { lo = FIT.grid[0]; hi = FIT.grid[FIT.grid.length - 1]; }
    else if (lxs.length) { lo = Math.min.apply(null, lxs) - 0.6; hi = Math.max.apply(null, lxs) + 0.6; }
    else { lo = -1; hi = 3; }
    if (hi - lo < 0.5) { lo -= 0.5; hi += 0.5; }

    var ys = pts.map(function (p) { return p.y; });
    var ymin = Math.min.apply(null, ys.concat([0])) - 4;
    var ymax = Math.max.apply(null, ys.concat([100])) + 6;
    if (fitOk) { ymin = Math.min(ymin, FIT.bottom - 4); ymax = Math.max(ymax, FIT.top + 6); }

    var X = function (lx) { return L + (lx - lo) / (hi - lo) * pw; };
    var Y = function (v) { return T + ph - (v - ymin) / (ymax - ymin) * ph; };
    PLOT = { L: L, T: T, pw: pw, ph: ph, lo: lo, hi: hi, ymin: ymin, ymax: ymax, X: X, Y: Y };

    ctx.font = '10px ' + css('--font-mono');
    ctx.lineWidth = 1;

    /* y grid — the step has to stay coarse enough that a typo like 1e9 in a
       response cell cannot spin this loop millions of times and hang the tab */
    var yr = ymax - ymin;
    var step = yr > 160 ? 50 : 25;
    if (yr / step > 12) step = niceStep(yr, 8);
    ctx.textAlign = 'right'; ctx.textBaseline = 'middle';
    for (var v = Math.ceil(ymin / step) * step; v <= ymax; v += step) {
      var yy = Math.round(Y(v)) + 0.5;
      ctx.strokeStyle = line; ctx.beginPath();
      ctx.moveTo(L, yy); ctx.lineTo(L + pw, yy); ctx.stroke();
      ctx.fillStyle = muted; ctx.fillText(axisNum(v), L - 7, yy);
    }

    /* x grid: decades + minor ticks */
    ctx.textAlign = 'center'; ctx.textBaseline = 'top';
    var d0 = Math.floor(lo), d1 = Math.ceil(hi);
    var dstep = Math.max(1, Math.ceil((d1 - d0) / 10));   /* never more than ~11 decades */
    var minorLine = rgba(css('--line'), 0.55);
    for (var d = d0; d <= d1; d += dstep) {
      for (var m = 1; m < 10; m++) {
        if (m > 1 && dstep > 1) break;   /* minor ticks only while decades are readable */
        var lx = d + Math.log10(m);
        if (lx < lo || lx > hi) continue;
        var xx = Math.round(X(lx)) + 0.5;
        ctx.strokeStyle = m === 1 ? line : minorLine;
        ctx.beginPath();
        ctx.moveTo(xx, T); ctx.lineTo(xx, T + ph); ctx.stroke();
      }
      if (d < lo || d > hi) continue;
      var px = X(d);
      var val = Math.pow(10, d);
      var lab = (d >= 5 || d <= -4) ? '1e' + d
        : (val >= 1000 ? (val / 1000) + 'k' : (val >= 1 ? String(val) : String(+val.toFixed(4))));
      ctx.fillStyle = muted;
      if (px > L + 8 && px < L + pw - 8) ctx.fillText(lab, px, T + ph + 7);
    }

    /* bootstrap band */
    if (gridOk && BOOT && BOOT.bandOk && curveT >= 1) {
      ctx.fillStyle = rgba(accent, 0.16);
      ctx.beginPath();
      FIT.grid.forEach(function (g, i) {
        var xp = X(g), yp = Y(Math.min(ymax, Math.max(ymin, BOOT.band_hi[i])));
        if (i === 0) ctx.moveTo(xp, yp); else ctx.lineTo(xp, yp);
      });
      for (var i2 = FIT.grid.length - 1; i2 >= 0; i2--) {
        ctx.lineTo(X(FIT.grid[i2]), Y(Math.min(ymax, Math.max(ymin, BOOT.band_lo[i2]))));
      }
      ctx.closePath(); ctx.fill();
    }

    /* IC50 CI on the x axis */
    if (gridOk && BOOT && fin(BOOT.ic50_lo) && fin(BOOT.ic50_hi) && BOOT.ic50_lo > 0) {
      var a = X(Math.max(lo, Math.log10(BOOT.ic50_lo)));
      var b2 = X(Math.min(hi, Math.log10(BOOT.ic50_hi)));
      ctx.fillStyle = rgba(muted, 0.22);
      ctx.fillRect(a, T + ph - 12, Math.max(1, b2 - a), 12);
      ctx.strokeStyle = rgba(muted, 0.9);
      ctx.beginPath();
      ctx.moveTo(a, T + ph - 12); ctx.lineTo(a, T + ph);
      ctx.moveTo(b2, T + ph - 12); ctx.lineTo(b2, T + ph);
      ctx.stroke();
    }

    /* fitted curve */
    if (fitOk && gridOk) {
      var n = Math.max(2, Math.round(FIT.grid.length * curveT));
      ctx.strokeStyle = accent; ctx.lineWidth = 2; ctx.lineJoin = 'round';
      ctx.beginPath();
      for (var i3 = 0; i3 < n; i3++) {
        var xp3 = X(FIT.grid[i3]), yp3 = Y(FIT.curve[i3]);
        if (i3 === 0) ctx.moveTo(xp3, yp3); else ctx.lineTo(xp3, yp3);
      }
      ctx.stroke();
      ctx.lineWidth = 1;

      /* IC50 crosshair */
      if (curveT >= 1) {
        var mid = (FIT.top + FIT.bottom) / 2;
        var cx = X(FIT.logic50), cy = Y(mid);
        ctx.save();
        ctx.setLineDash([4, 4]); ctx.strokeStyle = rgba(muted, 0.95);
        ctx.beginPath();
        ctx.moveTo(cx, cy); ctx.lineTo(cx, T + ph);
        ctx.moveTo(L, cy); ctx.lineTo(cx, cy);
        ctx.stroke();
        ctx.restore();
        ctx.fillStyle = surface; ctx.strokeStyle = accent; ctx.lineWidth = 2;
        ctx.beginPath(); ctx.arc(cx, cy, 4.5, 0, 6.2832); ctx.fill(); ctx.stroke();
        ctx.lineWidth = 1;
        ctx.fillStyle = muted; ctx.textAlign = 'left'; ctx.textBaseline = 'bottom';
        var lbl = 'IC50 ' + sig(FIT.ic50, 3) + ' nM';
        var tw = ctx.measureText(lbl).width;
        ctx.fillText(lbl, Math.min(cx + 7, L + pw - tw - 2), Math.max(T + 11, cy - 7));
      }
    }

    /* points */
    pts.forEach(function (p) {
      var xp = X(Math.log10(p.x)), yp = Y(p.y);
      var on = p.i === hoverIdx;
      ctx.beginPath(); ctx.arc(xp, yp, on ? 6.5 : 4.2, 0, 6.2832);
      ctx.fillStyle = on ? accent : ink;
      ctx.strokeStyle = surface; ctx.lineWidth = 2;
      ctx.fill(); ctx.stroke();
      ctx.lineWidth = 1;
    });

    /* axis titles */
    ctx.fillStyle = muted; ctx.textAlign = 'center'; ctx.textBaseline = 'bottom';
    ctx.fillText('concentration nM (log)', L + pw / 2, h - 1);
    ctx.save();
    ctx.translate(9, T + ph / 2); ctx.rotate(-Math.PI / 2);
    ctx.textBaseline = 'top'; ctx.fillText('% response', 0, 0);
    ctx.restore();
  }

  /* hover */
  function nearest(ev) {
    if (!PLOT) return -1;
    var r = cv.getBoundingClientRect();
    var mx = ev.clientX - r.left, my = ev.clientY - r.top;
    var best = -1, bd = 26 * 26;
    validPoints().forEach(function (p) {
      var dx = PLOT.X(Math.log10(p.x)) - mx, dy = PLOT.Y(p.y) - my;
      var d = dx * dx + dy * dy;
      if (d < bd) { bd = d; best = p.i; }
    });
    return best;
  }
  cv.addEventListener('pointermove', function (e) {
    var i = nearest(e);
    if (i !== hoverIdx) { hoverIdx = i; syncHover(); }
  });
  cv.addEventListener('pointerleave', function () { hoverIdx = -1; syncHover(); });

  function hoverText() {
    var pts = validPoints();
    var k = -1;
    pts.forEach(function (p, j) { if (p.i === hoverIdx) k = j; });
    Array.prototype.forEach.call($('tbody').children, function (tr, i) {
      tr.classList.toggle('hi', i === hoverIdx);
    });
    var el = $('hover');
    if (k < 0) {
      el.textContent = 'hover the chart, or focus a row below, to inspect a point';
    } else {
      var p = pts[k];
      var j = FIT && FIT.idx ? FIT.idx.indexOf(p.i) : -1;
      var r = j >= 0 && FIT.resid && fin(FIT.resid[j]) ? FIT.resid[j] : null;
      el.innerHTML = 'point <b>' + (p.i + 1) + '</b> · <b>' + sig(p.x, 3) + '</b> nM → <b>'
        + num(p.y, 1) + '</b> % response'
        + (r == null ? '' : ' · residual <b>' + (r >= 0 ? '+' : '') + num(r, 2) + '</b>');
    }
  }
  function syncHover() { hoverText(); draw(); }

  /* ---------------- readouts ---------------- */
  var live = document.createElement('p');
  live.className = 'sr'; live.setAttribute('aria-live', 'polite');
  document.body.appendChild(live);

  function chip(txt, cls) {
    var s = document.createElement('span');
    s.className = 'chip' + (cls ? ' ' + cls : '');
    s.innerHTML = txt;
    return s;
  }

  /* when a fit dies, the previous numbers and the previous curve are no longer
     about the data on screen — blank them rather than let them read as a result */
  function clearReadouts(why) {
    ['oIc50', 'oHill', 'oTop', 'oBot', 'oR2'].forEach(function (k) { $(k).textContent = '—'; });
    $('oIc50ci').textContent = '95% CI —';
    $('oHillci').textContent = '95% CI —';
    $('oRmse').textContent = 'RMSE —';
    $('fitTime').textContent = '—';
    $('chips').textContent = '';
    $('chips').appendChild(chip(why, 'bad'));
    Array.prototype.forEach.call($('tbody').querySelectorAll('[data-resid]'),
      function (td) { td.textContent = '·'; });
    cv.setAttribute('aria-label', 'Dose-response plot: ' + validPoints().length
      + ' measured points on a log concentration axis, no fitted curve.');
    hoverText();
  }

  function updateReadouts(progress) {
    if (!FIT) return;
    $('oIc50').innerHTML = sig(FIT.ic50, 3) + ' <small>nM</small>';
    $('oHill').textContent = num(FIT.hill, 2);
    $('oTop').textContent = num(FIT.top, 1);
    $('oBot').textContent = num(FIT.bottom, 1);
    $('oR2').textContent = num(FIT.r2, 4);
    $('oRmse').textContent = 'RMSE ' + num(FIT.rmse, 2);
    $('fitTime').textContent = num(FIT.ms, 1) + ' ms · ' + FIT.nev + ' evals';

    if (BOOT) {
      $('oIc50ci').textContent = '95% CI ' + sig(BOOT.ic50_lo, 3) + '–' + sig(BOOT.ic50_hi, 3);
      /* the hill slope is bounded at 0.05..12 by the penalty; say so instead of
         printing the bound as if it were an estimate */
      var hl = !fin(BOOT.hill_lo) ? '—' : BOOT.hill_lo <= 0.06 ? 'under 0.05' : num(BOOT.hill_lo, 2);
      var hh = !fin(BOOT.hill_hi) ? '—' : BOOT.hill_hi >= 11.5 ? 'over 12' : num(BOOT.hill_hi, 2);
      var atBound = hl === 'under 0.05' || hh === 'over 12';
      $('oHillci').textContent = '95% CI ' + hl + (atBound ? ' to ' : '–') + hh;
    } else {
      $('oIc50ci').textContent = progress ? '95% CI · ' + progress : '95% CI —';
      $('oHillci').textContent = '95% CI —';
    }

    var c = $('chips');
    c.textContent = '';
    c.appendChild(chip(FIT.converged ? '✓ simplex converged' : '⚠ hit iteration cap',
      FIT.converged ? 'ok' : 'bad'));
    c.appendChild(chip('iterations <b>' + FIT.iters + '</b>'));
    c.appendChild(chip('dof <b>' + FIT.dof + '</b>'));
    if (BOOT) {
      c.appendChild(chip('bootstrap <b>' + BOOT.n + '</b> refits · ' + num(BOOT.ms, 0) + ' ms', 'ok'));
      var w = BOOT.ic50_hi / BOOT.ic50_lo;
      c.appendChild(chip('CI width <b>' + num(w, 1) + '×</b>', w > 4 ? 'bad' : ''));
      if (BOOT.nonconv) c.appendChild(chip('<b>' + BOOT.nonconv + '</b> refits hit the cap', 'bad'));
    } else if (progress) {
      c.appendChild(chip('bootstrapping <b>' + progress + '</b>'));
    }

    var cells = $('tbody').querySelectorAll('[data-resid]');
    Array.prototype.forEach.call(cells, function (td) { td.textContent = '·'; });
    /* residuals belong to the rows that were fitted, not to whatever the table
       holds now — rows can be edited or deleted while a fit is in flight */
    (FIT.idx || []).forEach(function (row, j) {
      var td = cells[row], v = FIT.resid[j];
      if (td && fin(v)) td.textContent = (v >= 0 ? '+' : '') + num(v, 1);
    });

    hoverText();
    cv.setAttribute('aria-label',
      'Dose-response plot: ' + validPoints().length + ' measured points on a log concentration axis '
      + 'with a fitted Hill curve. IC50 ' + sig(FIT.ic50, 3) + ' nanomolar, hill slope '
      + num(FIT.hill, 2) + ', top ' + num(FIT.top, 1) + ' percent, bottom ' + num(FIT.bottom, 1)
      + ' percent, R squared ' + num(FIT.r2, 4) + '.');

    if (BOOT) {
      live.textContent = 'Fit complete. IC50 ' + sig(FIT.ic50, 3) + ' nanomolar, 95% confidence interval '
        + sig(BOOT.ic50_lo, 3) + ' to ' + sig(BOOT.ic50_hi, 3) + '. Hill slope ' + num(FIT.hill, 2)
        + '. R squared ' + num(FIT.r2, 4) + '.';
    }
  }

  var animId = 0;
  function animateCurve() {
    var id = ++animId;
    if (reduceMotion) { curveT = 1; draw(); return; }
    curveT = 0;
    var t0 = performance.now();
    (function step() {
      if (id !== animId) return;
      var k = Math.min(1, (performance.now() - t0) / 420);
      curveT = k * k * (3 - 2 * k);
      draw();
      if (k < 1) requestAnimationFrame(step);
      else { curveT = 1; draw(); }
    })();
  }

  /* ---------------- fit pipeline ---------------- */
  var timer = null;
  function scheduleFit(delay) {
    clearTimeout(timer);
    timer = setTimeout(runFit, delay || 0);
  }
  function setBusy(v) {
    busy = v;
    $('refit').disabled = v;
    $('refit').textContent = v ? 'fitting…' : 'refit ▸';
  }
  function frame() {
    return new Promise(function (r) { requestAnimationFrame(function () { setTimeout(r, 0); }); });
  }

  function errText(err) {
    if (!err) return 'unknown error';
    var m = err.message || String(err);
    return m.split('\n').filter(function (l) { return l.trim(); }).pop() || m;
  }

  async function runFit() {
    /* the chart still has to reflect the edit even when there is nothing to fit with */
    if (!py) { draw(); return; }
    var pts = validPoints();
    if (pts.length < 4) {
      /* cancel whatever is in flight and hand the controls back, otherwise the
         superseded run never releases them and refit sits at "fitting…" */
      runToken++;
      setBusy(false);
      draw();
      flashNote('need at least 4 valid points (concentration > 0)');
      return;
    }
    var token = ++runToken;
    setBusy(true);
    BOOT = null;

    try {
      FIT = JSON.parse(pyfn.fit(JSON.stringify({
        x: pts.map(function (p) { return p.x; }),
        y: pts.map(function (p) { return p.y; }),
        fix_bottom: fixBottom
      })));
      FIT.idx = pts.map(function (p) { return p.i; });
    } catch (err) {
      /* a superseded run must not hand the controls back, or report — the newer one owns them */
      if (token !== runToken) return;
      setBusy(false);
      FIT = null;
      clearReadouts('fit failed');
      flashNote('fit failed: ' + errText(err));
      draw();
      return;
    }
    if (token !== runToken) return;

    updateReadouts('0/' + bootN);
    animateCurve();

    try {
      pyfn.boot_start(bootN, 7);
      /* one chunk per frame. 100 keeps each blocking numpy pass around 100ms and
         leaves the counter something to say at 500 and 1000 resamples — a chunk of
         500 froze the tab for half a second at a time */
      var chunk = 100;
      for (var done = 0; done < bootN;) {
        await frame();
        if (token !== runToken) return;
        done = pyfn.boot_run(Math.min(chunk, bootN - done));
        updateReadouts(done + '/' + bootN);
      }
      BOOT = JSON.parse(pyfn.boot_finish());
      BOOT.bandOk = BOOT.band_lo.every(fin) && BOOT.band_hi.every(fin);
    } catch (err) {
      if (token !== runToken) return;
      BOOT = null;
      flashNote('bootstrap failed: ' + errText(err));
    }
    if (token !== runToken) return;
    curveT = 1;
    updateReadouts();
    draw();
    setBusy(false);
  }

  /* ---------------- controls ---------------- */
  $('refit').addEventListener('click', function () { scheduleFit(0); });
  $('addPt').addEventListener('click', function () {
    var last = DATA[DATA.length - 1];
    var lx = last && isFinite(last.x) && last.x > 0 ? last.x : 1;
    var ly = last && isFinite(last.y) ? last.y : 50;
    DATA.push({ x: +(lx * 3).toPrecision(3), y: Math.max(0, +(ly * 0.6).toFixed(1)) });
    activePreset = null; renderTable(); markPresets(); scheduleFit(0);
    var rows = $('tbody').children;
    rows[rows.length - 1].querySelector('input').focus();
  });
  $('fixBottom').addEventListener('change', function (e) {
    fixBottom = e.target.checked; scheduleFit(0);
  });
  $('bootN').addEventListener('change', function (e) {
    bootN = parseInt(e.target.value, 10); scheduleFit(0);
  });

  /* ---------------- theme + resize ---------------- */
  window.addEventListener('message', function (e) {
    var d = e.data;
    if (d && d.type === 'sloprun:theme' && (d.theme === 'light' || d.theme === 'dark')) {
      document.documentElement.setAttribute('data-theme', d.theme);
      draw();
    }
  });
  var dark = matchMedia('(prefers-color-scheme: dark)');
  if (dark.addEventListener) dark.addEventListener('change', function () { draw(); });
  else if (dark.addListener) dark.addListener(function () { draw(); });
  /* the canvas paints its own background and axes, so it has to repaint whenever the
     tokens change — including a host that sets data-theme on <html> directly */
  new MutationObserver(function () { draw(); })
    .observe(document.documentElement, { attributes: true, attributeFilter: ['data-theme'] });

  /* only width drives the canvas geometry; reacting to the height this very function
     sets is how a ResizeObserver ends up looping */
  var lastW = -1;
  var ro = new ResizeObserver(function (entries) {
    var w = Math.round(entries[0].contentRect.width);
    if (w === lastW) return;
    lastW = w;
    draw();
  });
  ro.observe(cv.parentElement);

  /* ---------------- boot ---------------- */
  var bootPct = 0, creepTimer = 0;
  function paintBoot(done) {
    $('bootPct').textContent = Math.round(bootPct) + '%';
    $('bootBar').firstElementChild.style.width = bootPct + '%';
    $('bootBar').classList.toggle('done', !!done);
  }
  function setStage(txt, pct, done) {
    clearInterval(creepTimer); creepTimer = 0;
    $('bootStage').textContent = txt;
    bootPct = Math.max(bootPct, pct);
    paintBoot(done);
  }
  /* loadPyodide and loadPackage report nothing while they run, so the bar eases
     toward the next milestone instead of sitting frozen on a cold cache */
  function creep(cap) {
    clearInterval(creepTimer);
    creepTimer = setInterval(function () {
      if (bootPct >= cap - 0.4) { clearInterval(creepTimer); creepTimer = 0; return; }
      bootPct += (cap - bootPct) * 0.07;
      paintBoot(false);
    }, 200);
  }
  function bootFail(msg) {
    clearInterval(creepTimer); creepTimer = 0;
    var b = $('boot');
    b.classList.add('fail');
    $('bootStage').textContent = 'python runtime unavailable';
    $('bootPct').textContent = '—';
    $('bootBar').firstElementChild.style.width = '100%';
    $('bootBar').firstElementChild.style.background = 'var(--danger)';
    $('bootSub').innerHTML = msg;
    $('refit').disabled = true;
    $('refit').textContent = 'no runtime';
    /* these two only ever change the fit — with no fitter they are decoration */
    $('fixBottom').disabled = true;
    $('bootN').disabled = true;
    $('chips').textContent = '';
    $('chips').appendChild(chip('runtime <b>unavailable</b>', 'bad'));
    $('note').textContent = 'The data below is real and the plot is live — only the Python fit is missing.';
    noteIdle = 'Edit any cell — the plot follows along, but fitting needs the Python runtime.';
    $('dataNote').textContent = noteIdle;
    draw();
  }
  function loadScript(src) {
    return new Promise(function (res, rej) {
      var s = document.createElement('script');
      s.src = src;
      s.onload = res;
      s.onerror = function () { rej(new Error('could not load ' + src)); };
      document.head.appendChild(s);
    });
  }

  async function boot() {
    /* first paint is the demo itself: real assay data in the table, real points
       on the chart. Only the fit is gated on the runtime landing. */
    renderTable(); buildPresets();
    $('note').textContent = PRESETS[0].note;
    $('refit').disabled = true;
    $('refit').textContent = 'waiting…';
    $('chips').appendChild(chip('runtime <b>loading…</b>'));
    draw();
    try {
      setStage('fetching python runtime…', 8); creep(30);
      await loadScript('/vendor/pyodide/pyodide.js');
      if (typeof loadPyodide !== 'function') throw new Error('loadPyodide missing');
      setStage('starting CPython (wasm)…', 34); creep(58);
      await frame();
      py = await loadPyodide({ indexURL: '/vendor/pyodide/' });
      setStage('loading numpy…', 62); creep(80);
      await frame();
      await py.loadPackage('numpy');
      setStage('compiling the fitter…', 84);
      await frame();
      py.runPython($('pysrc').textContent);
      ['fit', 'boot_start', 'boot_run', 'boot_finish'].forEach(function (k) {
        pyfn[k] = py.globals.get(k);
      });
      $('engine').textContent = 'python: ' + py.runPython(
        'import sys, numpy; sys.version.split()[0] + " · numpy " + numpy.__version__');
    } catch (err) {
      bootFail('Pyodide was not found at <span style="font-family:var(--font-mono)">/vendor/pyodide/</span>. '
        + 'On sloprun.dev it is served from the platform bundle; opened as a bare local file there is no '
        + 'runtime to fit with, and this demo never reaches out to a CDN.');
      return;
    }
    setStage('✓ runtime ready', 100, true);
    $('refit').disabled = false;
    $('refit').textContent = 'refit ▸';
    draw();
    await runFit();
    setTimeout(function () { $('boot').hidden = true; }, 700);
  }

  boot();
})();
</script>
</body>
</html>