I posted some time ago asking whether the new POP-5 Beta
Basecaller[1], etc. was good or not. George Grills wrote me and said
it solved some of the POP-5 mobility shift problems. So I tried it.
It isn't just a basecaller. It includes a basecaller, a mobility
file and a new dll. Since I had a number of runs still in my 3700
reduced read data base, I changed the mobility files and reextracted
the runs. This was an agonizingly slow and inefficient process.
Disappointingly many of the runs were not reextracted correctly and
gave no useful data. But here are the ones that did work. I used phred
to give quality scores. These are the same runs, extracted first using
the old mobility file and then again with the new mobility file:
H I G H Q U A L I T Y B A S E S
old mobility new mobility improvement
run mean median mean median mean median
___ ______________ ______________ ______________
473 537.3 584 572.4 622 35.1 38
487 512.5 643 541.2 661 28.7 18
486 513.9 637 537.8 657 23.9 20
475 584.9 632 615.6 667 30.7 35
482 270.6 330 292.3 348 21.7 18
483 480 618 505.2 645 25.2 27
454 427.7 504 454 529 26.3 25
453 564.9 635 594.1 656 29.2 21
452 595.5 644 631.1 685 35.6 41
456 533.6 633 555.3 654 21.7 21
455 531.6 638 456.9 527 -74.7 -111
460 488 543 515.5 605 27.5 62
462 553.9 623 585.9 659 32 36
461 556.7 636 492.1 535 -64.6 -101
467 297.8 377 316.6 389 18.8 12
478 382 477 408.3 473 26.3 -4
468 282.6 338 305.2 350 22.6 12
479 445.8 518 151.6 302 -294.2 -216
469 228.2 306 246.8 378 18.6 72
480 399.5 469 422.1 508 22.6 39
481 396.9 508 417.4 540 20.5 32
485 114 218 128.2 204 14.2 -14
488 573.6 635 596 651 22.4 16
489 128.5 244 142 250 13.5 6
451 559.4 618 588.6 640 29.2 22
466 367.2 406 399.5 425 32.3 19
472 592 630 617.8 653 25.8 23
474 585.5 628 611.4 660 25.9 32
476 586.2 612 612.3 636 26.1 24
477 572.2 600 599.6 626 27.4 26
490 586.3 602 614.8 630 28.5 28
471 568.2 644 595.6 680 27.4 36
457 631 664 663.9 695 32.9 31
464 306.5 452 330.9 484 24.4 32
465 187.8 30 205.8 32 18 2
508 563.1 660 594.3 694 31.2 34
Conclusion? In most cases there was a substantial improvement in
number of phred q>20 bases. But in a few cases there was an alarming
decrease. Perhaps these were actually artifacts of the re-extraction
process. I haven't looked at the individual chromat to find out what
is going on.
I have done more than 100 runs since changing and the results have
be very good. Indeed it is not uncommon for me to get median numbers
of HQBs over 700. Our best read ever: 805 high quality bases. (Has
anyone else broken 800? We have only done so once--although HQBs in
the 790s appear from time to time.)
Do the increases in number of phred HQBs represent a true increase
in calling accuracy. I don't know.
Phillip San Miguel
Purdue Genomics Core Facility
[1] Download at:
http://www.appliedbiosystems.com/techsupp/swpps/3700sw.html
---
