Optimization of polymer replacement protocol for novel automated single-lane DNA sequencing instrument based on single photon detection

Olga Bilenko*, Dmitri N. Gavrilov, Boris Gorbovitski*, Mikhail Gouzman, Georgiy Gudkov, Vyacheslav Khozikov, Olga Kosobokova, Nadia Lifshitz*, Serge Luryi, Andrew Stepoukhovitch, Marina Tcherevishnick, Andrei Tsuprik, Georgiy Tyshko, and Vera Gorfinkel Department of Electrical and Computer Engineering, SUNY SB, Stony Brook, NY *BioPhotonics Corp. Stony Brook, NY

Presented at 16th International Symposium on Microscale Separations and Analysis (HPCE 2003), San Diego, January 17-22, 2003 To appear in Journal of Chromatography A (2003) Corresponding author: Olga Kosobokova Department of Electrical and Computer Engineering, SUNY SB, Stony Brook, NY, 11794-2350 Tel.: +631-632-1608 FAX: +631-632-5745 e-mail: okey@ece.sunysb.edu

Optimization of polymer replacement protocol for novel automated single-lane DNA sequencing instrument based on single photon detection
Olga Bilenko*, Dmitri N. Gavrilov, Boris Gorbovitski*, Mikhail Gouzman, Georgiy Gudkov, Vyacheslav Khozikov, Olga Kosobokova, Nadia Lifshitz*, Serge Luryi, Andrew Stepoukhovitch, Marina Tcherevishnick, Andrei Tsuprik, Georgiy Tyshko, and Vera Gorfinkel Department of Electrical and Computer Engineering, SUNY SB, Stony Brook, NY *BioPhotonics Corp. Stony Brook, NY

Abstract
An automated ultra-sensitive single-lane DNA sequencer based on single photon counting fluorescent detection technique has been developed and tested. The instrument is intended for individual researchers who carry out a limited number of sequencing experiments (about 106 bp/year, the size typical for small-scale research projects). Due to the use of single photon detection, sensitivity of the instrument is about 100 fold higher than in commercial ABI instruments. The sequencer has a flexible modular fiber-optic design and can be easily configured to operate with different laser sources, as well as various fiberized photodetectors. A novel automation concept providing the ability of the remote (Internet) access and control of the system has been implemented and tested for DNA sequencing and fragment analysis applications. Optimization of sample running protocol enabled up to 12 hours of continuous operation without polymer replacement for standard ABI sequencing chemistry and reagents.

List of keywords:
DNA sequencing, single photon detection, automated operation.

1. Introduction
Capillary electrophoresis is widely used for the high-throughput DNA sequencing. Modern commercial sequencing machines employ a multi-capillary format (ABI PRIZM-3100, CEQ-2000, ABI PRIZM-3700, MegaBACE). The highest number of lanes (384) is offered by the MegaBACE4000 sequencer. At the same time, several groups develop new high-throughput sequencing machines based on integrated-system technologies and prepare to overcome the 1,000 channel barrier [1-10]. By their technical principle, the fluorescence detection systems used in current DNA sequencing instruments are very similar to those developed in the mid 1980s-1990s. They excite fluorescence markers using a laser source and then capture fluorescence with an analog photo detector (either of PMT or CCD type). This signal is then digitized, transferred to a computer, and analyzed to determine the sequence. There are two major technical shortcomings common to all 4color machines: the relatively low detection sensitivity and the relatively narrow dynamic range, limited by the analog-to digital conversion circuitry. Much higher detection sensitivity and much greater dynamic range are needed to substantially reduce the sequencing costs and improve the data quality.
2

A number of extremely sensitive fluorescence detection techniques are available based on registering single photons. Commonly, they are referred to as the photon correlation techniques. Until very recently single-photon counting techniques were mostly used for specialized scientific applications, such as the detection of single fluorescent molecules [11-15]. Though very sensitive, these techniques are not widely used in commercial DNA sequencing. The two primary obstacles preventing the wider use of these techniques are their complexity and the high cost per channel. In our previous publication we described a new family of DNA sequencing instruments based on single photon detection [16]. In these instruments we have implemented high performance, low cost single photon detectors and demonstrated a 100 fold higher sensitivity as compared to commercial instruments [17]. In this paper we describe an automated ultra-sensitive single-lane DNA sequencer based on single photon counting which was developed at SUNY SB and Biophotonics Corporation (the work is supported by the NIH (NHGRI, NCI)). The instrument is intended for individual researchers who carry out a limited number of sequencing experiments (about 106 bp/year, the size typical for smallscale research projects). The sequencer has flexible modular fiber optic design and can be easily configured to operate with various laser sources, including Ar-ion laser (488 nm, 514 nm), Nd:YAG (532 nm) and red-infrared laser diodes as well as various fiberized photodetectors. A novel automation concept providing the ability of the remote (Internet) access and control of the system has been implemented and tested for DNA sequencing and fragment analysis applications. Optimization of sample running protocol enabled for up to 12 hours of continuous operation without polymer replacement for standard ABI sequencing chemistry and reagents.

2. Materials and methods

2.1. Automated DNA sequencer SBS-2002 Detailed description of the manual single-lane DNA sequencer based on single photon counting fluorescent detection technology (SBS-2000 model) is given in [16,17]. During the first round of testing of the SBS-2000, we found a number of bottlenecks in our data acquisition system. Based on system analysis, we have formulated specific requirements to the sequencer automation system and developed a general automation concept capable of addressing the system bottlenecks. The concept includes three basic principles: (i) The system should allow creating the software package which would be independent on the upgrades in both PCs and their operating systems; (ii) Provide both the soft- and the hardware platforms for broad further increase in automation, including remote receiving of the sequencing data as well as remote control of the sequencers; (iii) Provide means for a packet-switched method of data transmission and processing during long-term operation (several days) with an automated recording of all results in the appropriate databases. In the present paper we describe an automated version of the instrument (SBS-2002 model), based on the above principles. A photograph and block-diagram of the automated SBS-2002 sequencer are shown in Fig.1 and Fig.2. The instrument comprises three modules connected via optical fibers. Labeled DNA sample undergoes separation in a single-capillary Sequencing Module. At the detection end of the capillary a laser beam from the Laser Module delivered via 60m fiber excites the fluorescence of DNA zones. The fluorescence is collected by 200m fiber-receiver and delivered to the Photoreceiving Module, and recorded by a computer. The automated system is divided into Sequencer and Controller parts. The Sequencer and the Controller are two computers separated in space. They communicate with each other via Ethernet using standard TCP/IP protocol. While the Sequencing Module, Laser Module, and the Photoreceiving Module are located on the Sequencer's end of the

transmission line, data collection, visualization, processing, and automated control of the process execution are performed on the Controller's end. The Sequencer is based on PC-104 computer board, which communicates with an Automation Control Module using an RS-232 Serial Interface, and with a Photodetector Module using IEEE-1284 Parallel Interface. The Automation Control Module, based on a Motorola M68000 processor, is built into the Sequencing Module. It controls the operation of two drivers, one electromechanical, the other high-voltage. The electromechanical driver is responsible for the operation of motor drivers that control and record positions of step motors that independently set the vertical and horizontal positions of the injection and outlet vials. It also includes XY position sensors that register the position of the injection and outlet vials and send this information back to the Automation Control Module. To automatically regulate the voltage in the injection and the sequencing regimes and to record current flowing through the capillary, we have designed an automated High Voltage Source which is connected to the Automation Control Module via AD/DA converters. The Automation Control Module accepts commands from the Sequencer (PC-104), controls their execution, and reports the results to the Sequencer. The command set includes control commands for motions of the sample trays and setting the voltage applied across the capillary. It also includes a number of request commands about the sample tray position, voltage across the capillary and the current in it.
Capillary Receiving fiber Illumination fiber Buffer
Reading head

Singlephoton detector

Laser

DNA sample

Computer

Figure 1. Automated DNA sequencer SBS-2002. The instrument comprises three modules connected via optical fibers. Labeled DNA sample undergoes separation in a single-capillary Sequencing Module. At the detection end of the capillary a laser beam from the Laser Module delivered via 60m fiber excites the fluorescence of DNA zones. The fluorescence is collected by 200 m fiber and delivered to the Photo-detection Module, and recorded by a computer.

1''

In the Photoreceiving Module, a single photon PMT detector (H-6240-02 or H-7467 from Hamamatsu) receives fluorescent signals via an optical fiber from the Sequencing Module, amplifies current pulses produced by single photons, converts them into a standard TTL format and transfers the TTL signals to a photon counter. The photon counter based on a field programmed gate array (FPGA), registers the photon pulses and transmits the recorded data to the Sequencer (PC104), which in turn retransmits the data to the Controller PC via Ethernet. The Sequencer also receives and dispatches the commands from the Controller, communicating with the Automation Control Module and responding to the Controller. The Controller is a PC with our original software (Fig. 3). The software includes tools for the selection and editing protocols for sequencing runs. The programming of a series of runs is performed by the selection of positions of the sample tubes for pre-run, injection and sequencing phases of each run, as well as setting the time of each phase and voltage applied across the capillary

during each phase. After the series of runs is started by an operator, the Controller controls the Sequencer by sending the appropriate commands. The arriving data stream is visualized and recorded. The Controller software produces a separate raw data file for every sequencing run in the series. Execution of runs can be interrupted and resumed by the operator at any time. The Controller software is currently P h o to d e te c tio n m o d u le S e q u e n c in g m o d u le implemented using Visual C++ 6.0 Laser Module MFC library. The availability of a D N A s a m p le Laser Photo-detection Module standard TCP/IP interface for a variety lo a d in g u n it Photo-detection module of computer platforms allows an P o ly m e r O p tic a l s y s te m re p la c e m e n t implementation of the Controller u n it F ilte r m o d u le software based one of the crossT h e rm o c o n tro l u n it platform languages, such as JAVA. The M u lti-cphoton l hanne H ig h V o lta g e Single next step of the software development s in g le p h o to n s u p p ly detector d e te c to r will be implementation of the Controller software that will control P h o to n c o u n te r and collect data from multiple Automated DNA Sequencers. Figure 2. Block-diagram of the automated DNA sequencer. Additional software has been created Polymer replacement and thermo-control units are under that allows not only data collection, but construction. also control of the SBS-2002 instrument over the Ethernet.
CONTROLLER (Client, PC) Multi-capillary readinghead Capillary reading head Control Module (Motorola 68000) EHTERNET SEQUENCER (Server, PC 104)

a)

b)

Figure 3. Controller software: a) tubes position selection tool (upper panel), sequencing protocol selection tool (middle panel), data recording and visualization window (lower panel); b) base-calling and fragment analysis software. 2.2. Sequencing Reagents We have tested the instrument for DNA sequencing and fragment analysis applications using commercially available ABI four-color sequencing standards labeled with BigDye dye set (BigDyeTM Terminator Sequencing Standard, PN-4304154) and GenePrint PowerPlexTM 16 System (PN DC6531) samples (Internal Lane Standard 600 (ILS 600) and Allelic Ladder Mix) from Promega. All runs were carried out at room temperature. We used fused silica uncoated capillaries from PolyMacro, variable length (40-55 cm), variable ID (50 and 75 m). The detection window 1cm long was formed by removing the polyamide coating at 5.5 cm from the anode end of the capillary. The ABI consumables used in our experiments were ABI running buffer (PN 402824) and polymers POP-4 (PN 4028380) and POP-5 (PN 4313087) for uncoated capillaries. Filling the capillary with separation polymer was done using our external home-built filling station. A short

(usually 10-15 min) cleaning pre-run was conducted each time after fresh polymer loading. DNA samples were processed according to the suppliers' protocols and introduced into the capillary by electrokinetic injection. Optimal injection voltage and duration were determined to be 3 kV, 40 s for Sequencing Standard runs in 55-cm long, 75um ID capillaries and 15kV, 10s for fragment analysis in 40-cm long, 50um ID capillaries. 2.3. Optimization of the separation protocol In our previous work [18] we have studied the formation and properties of the anomalous resistive region which appears at the anode end of the capillary during the electrophoretic separation and interferes with a normal separation process in POP-5 and POP-4 polymers. This effect could be eliminated by adding a certain amount of separation polymer to the anode buffer vial for a single sequencing run. In the course of the present project we have defined optimum experimental conditions enabling for continuous multiple runs without polymer a) 1st run 400 bp replacement in the capillary. 6x10 Various mixtures of the POP-5 polymer and the ABI-310 running 0 4000s 6000s 8000s buffer (25%, 50%, 75%, 100% 2nd run polymer addition) were placed at the 2x10 outlet (anode) vial. A series of several consecutive runs were performed for 0 each of the above preceded by an 4000s 6000s 8000s initial 10-15 min cleaning pre-run. Time, sec Pre-run and separation were 1st run 400 bp b) conducted at 200 V/cm. We found 1x10 that the composition of 75% POP-5 concentration in the anode vial, 0 2000s 4000s 6000s 8000s allowed up to 12 hours of continuous 2nd run operation. Fig. 4 shows typical 400 bp 1x10 electropherograms for first and second consecutive runs (raw data, 0 baseline removed) obtained with pure 2000s 4000s 6000s 8000s buffer in both vials and with 75% Time, sec POP-5 content in the anode vial. For the POP-4 polymer used as a Figure 4. Optimization of sequencing protocol for DNA separation in POP-5 polymer: two consecutive runs with a) ABIseparation medium for fragment 310 running buffer in both (inlet and outlet) vials and b) 75% POPanalysis tests, the optimum 5 added to the outlet (anode) vial. Run conditions: 3 kV, 40 s composition in the output vial was 1:1 electrokinetic injection in 55-cm, 75 ID uncoated capillary, of the POP-4 and the ABI-310 buffer. separation voltage corresponding to 200 V/cm.
Fluorescence intensity, c/s
5 5

Fluorescence intensity, c/s

2.4. Preparation of sequencing traces for processing by PHRED package Algorithms used in PHRED for base-calling and quality values assignment are discussed in detail in [19,20]. A detailed description of the preprocessing procedures used for conversion of our sequencing data in the form accepted by PHRED is given in [16]. The procedure is a two-pass algorithm and includes the steps of data preparation, base-line removal, cross-talk filtering, peak smoothing, channel intensity equalization, peak detection, peak width equalization, mobility shift correction and data decimation. The procedure also includes adaptive features supported by peak

width estimation and cross-talk matrix adjustment steps. The estimates of instant peak width and adjusted cross-talk matrix are used for fine tuning of the processing algorithms during the second pass. The result of the conversion is saved in SCF format.

3. Results
3.1. DNA sequencing tests
1.5x10 1.0x10 5.0x10 0.0 1.0x10
6 6

1st run

2000s

4000s

6000s

8000s

2nd run

Fluorescence intensity, counts/sec

5.0x10 0.0

7.5x10 5.0x10 2.5x10 0.0 5.0x10

2000s

4000s

6000s

8000s

We have found that at the optimal concentration of POP-5 polymer in the anode vial (75%) the capillary can be used several times without refilling with fresh polymer. We were able to carry out up to 5 consecutive separations of DNA Sequence Standard (single run duration ~2.5 hours, total run time ~12 hours) without noticeable deterioration of the separation
60

3rd run

2000s

4000s

6000s

8000s

50 Average PHRED score

2.5x10 0.0

4th run

40 30 20 10 0

5.0x10

2000s

4000s

6000s

8000s

5th run

2.5x10 0.0

2000s

4000s

6000s

8000s

Time, sec

100

200

300

400

500

Base position, bp

Figure 5. Typical 5-run BigDyeTM Terminator Sequencing

Standard (ABI, PN 4304154) separation series in POP-5 (raw data). Run conditions: 3 kV, 40 s electrokinetic injection in 55-cm, 75 ID uncoated capillary, separation voltage corresponding to 200 V/cm.

Figure 6. Average PHRED score plot for a typical 5-run BigDyeTM Terminator Sequencing Standard (PN 4304154) separation series in POP-5 showing an average read length of >400 bp (at PHRED score 20).

quality. A general view of a typical 5-run series is presented in Fig. 5 (raw data), an average PHRED score plot for the given series is presented in Fig 6. The run model was tested on about 100 runs with read lengths of >400 bp (at PHRED score 20) and a maximum number of visually distinguished peaks up to 550. We found that the separation quality increases from run to run within first 5 runs in a series (Fig. 7, 100 to 150 bases and 400 to 450 bases processed fragments from the 1st and 5th runs) and then quickly degrades during the next run (data not shown). As it can be seen, each consecutive run is slowed down compared to the previous one. This behavior suggests certain improvements for the run model, such as an optimization of the pre-run time to achieve better separation in the very first run while still maintain high quality for up to 5 separations. Further

improvement of the separation quality as well as increase of the continuous operation time can be achieved by varying a polymer content in the outlet vial from run to run, which is currently under investigation.

a)

C CT T G A ACA A TG C GC C GT CAT GC T T C T T T T GC C T C CC GC T GC T CCA G AA AG

b)

CC T TG A AC AA TG C GC C GT CAT GC T T C T T T T GC C T C CC GC T GC T CCA G AA AG

100 bp

150 bp

100 bp

150 bp

G GC G AT CA ATA T TC C AT A A G AT G AT G GT T GC T CA G AGG CA G A G AA G A A GC G

G GC G A TC A A TA T TCC AT A AGGC ATG AT G GT T GC T CA G AGG CA GGA G A G A A A

400 bp

450 bp

400 bp

450 bp

Figure 7. Four-color DNA sequencing traces processed by PHRED. Resulting data from 100 to 150 bases and 400 to
450 bases is displayed for the first (a) and the fifth (b) runs of a typical 5-run series.

3.2. Fragment analysis tests

1x10
6

1st run

0 5x10
5

2nd run

Fluorescence intensity, counts/sec

5x10

3rd run

0 5x10
5

4th run

0 5x10
5

5th run

0 5x10
5

6th run

0 5x10
5

7th run

0 5x10
5

We have carried out about 100 test runs using POP-4 polymer as a separation medium in order to test fragment analysis application of the instrument. A general view of a typical 12-run series (ILS-600 and Allelic Ladder Mix sample, single run duration 35 min, total run time ~7 hours) is shown in Fig. 8 (raw data, baseline removed). As opposed to the preceding section, the run duration is decreased with time. Enlarged fragments view of the 1st, 6th and 12th runs (Fig. 9) does not show any degradation of the separation quality. Figure 8. Typical 12-run GenePrint PowerPlexTM
16 System (Promega, PN DC6531) samples (Internal Lane Standard 600 (ILS 600) and Allelic Ladder Mix) separation series in POP-4. Run conditions: 15 kV, 10 s electrokinetic injection in 40-cm, 50 ID uncoated capillary, separation voltage 15 kV.

8th run

0 5x10
5

10th run

0 3x10
5

12th run

1000s

1500s

2000s

Time, sec

4. Concluding remarks
We have described the operational principles, design and testing results of the automated DNA sequencing instrument based on single-photon counting. A novel automation concept was implemented in the instrument, providing the ability of the remote (Internet) access and control of the system operation. The proposed system provides a number of advantages. It allows the use of a variety of software and computer platforms since their communication is supported by the most universal TCP/IP protocol. For example, for the Sequencer, one can use any single board PC (PC-104, Ubicom, Microchip, etc.) and for the Controller any type of computer software and hardware. Another advantage is the ability of the remote access and control of the system, when the sequencing readout and control can be done from a remote terminal. The proposed automation concept is entirely applicable to multi-lane instruments. Considering the single-lane instrument to be a good testing ground for a new concept of the sequencer automation, we are presently working on introducing the automation into a multilane sequencer. The automated sequencer 2x10 1st run CRX TMR has been tested for DNA sequencing fluorescein and fragment analysis applications. 1x10 JOE Optimization of sample running protocol enabled for up to 12 hours 0 of continuous operation without 1000s 1100s 1200s 1300s polymer replacement for standard 8x10 ABI sequencing chemistry and 6th run reagents. Using the same capillary 4x10 for multiple runs without changing the separation medium between runs 0 reduces the polymer consumption 1000s 1100s 1200s 1300s and decreases total run time (since 4x10 only one pre-run is needed in this 12th run case) and, therefore, increases the instrument throughput. Depending 2x10 on the run duration this increase can be rather high. For example, typical 0 1000s 1100s 1200s 1300s run model for fragment analysis on Time, sec SB Sequencer consists of 10-minute pre-run and 35-minute run, that is Figure 9. Enlarged raw data (baseline removed) fragments of the 7.5 hours for 10-run series (not electropherograms obtained for ILS 600 + Allelic Ladder Mix including fresh polymer filling and samples during a typical 12-run separation series. Fluorescein, JOE capillary installation time). At the and TMR-labeled peaks corresponding to Allelic Ladder Mix and CRX-labeled peaks corresponding to ILS 600 are indicated. same time such a series would take only 6 hours when executed without polymer replacement, which immediately results in 20% increase of the instrument throughput. The modular architecture provides a tremendous flexibility in the design and the assembly of instruments for various applications, ranging vastly in their sensitivity, throughput, size and cost. We believe that the high sensitivity and dynamic range of our compact affordable automated DNA sequencer that can be installed on a tabletop in a laboratory and connected to the Internet will make it very useful for an individual researcher.
6 6 5 5 5 5

Fluorescence intensity, counts/sec

References
1. X. Lu and E. S. Yeung, Appl. Spectrosc. 49 (1995) 605. 2. Y. Wang, J.M. Wallin, J. Ju, G.F. Sensabaugh and R.A. Mathies, Electrophoresis 17 (1996) 1485. 3. I. Kheterpal, J.R. Scherer, S.M. Clark, A. Radhakrishnan, J. Ju, C.L. Ginther, G.F. Sensabaugh and R.A. Mathies, Electrophoresis 17 (1996) 1852. 4. Y. Wang, S.-H. Hung, J.F. Linn, G. Steiner, A.N. Glazer, D. Sidransky and R.A. Mathies, Electrophoresis 18 (1997) 1742. 5. J.R.Scherer, I. Kheterpal, A. Radhakrishnan, W.W. Ja, R.A. Mathies, Electrophoresis 20 (1999) 1508. 6. H.M. Pang, V. Pavski and E.S. Yeung, J. Biochem. Biophys. Meth. 41 (1999) 121. 7. S.X. Lu and E.S. Yeung, J. Chromatogr. A 853 (1999) 359. 8. J.Z. Zhang, K.O. Voss, D.F. Shaw, K.P. Roos, D.F. Lewis, J. Yan, R. Jiang, H. Ren, J.Y. Hou, Y. Fang, X. Puyang, H. Ahmadzadeh, and N.J. Dovichi, Nucleic Acids Research 27 (1999) E36. 9. H.John Crabtree, S.J. Bay, D.F. Lewis, L.D. Coulson, G. Fitzpatrick, D.J. Harrison, S.L. Delinger, J.Z. Zhang, and N.J. Dovichi, Electrophoresis 21 (2000) 1329. 10. S. Behr, M. Matzig, A. Levin, H. Eickhoff, C. Heller, Electrophoresis 20 (1999) 1492. 11. N.J. Dovichi, J.C. Martin, J.H. Jett, M. Trkula, and R.A. Keller, Analytical Chemistry 56 (1984) 348. 12. D.Y. Chen and N.J. Dovichi, Analytical Chemistry 68 (1996) 690. 13. S.A. Soper, L. M. Davis, E.B. Shera, J. Opt. Soc. Am. B 9 No. 10 (1992) 1761. 14. S.A. Soper, Q.L. Mattingly and P. Vegunta, Analytical Chemistry 6 (1993) 740. 15. P.M. Goodwin, R.J.Affleck, W.P.Ambros, J.N. Demos, J.H. Jett, J.C. Martin, L.J Reha-Krantz, D.J. Semin, J.A Schecker, M. Wu, R.A Keller, Experimental Technique of Physics 41 N2 (1995) 294. 16. L. Alaverdian, S. Alaverdian, O. Bilenko, I. Bogdanov, E. Filippova, D. Gavrilov, B. Gorbovitski, M. Gouzman, G. Gudkov, S. Domratchev, O. Kosobokova, N. Lifshitz, S. Luryi, V. Ruskovoloshin, A. Stepoukhovitch, M. Tcherevishnick, G. Tyshko, V. Gorfinkel, Electrophoresis 23 (2002) 2804. 17. O. Bilenko, D. Gavrilov, B. Gorbovitski, M. Gouzman, G. Gudkov, O. Kosobokova, N. Lifshitz, S. Luryi, A. Stepoukhovitch, M. Tcherevishnick, G. Tyshko, V. Gorfinkel, V. Balija, A. Bahret, R. Shah, L. King, L. Speigel, W.R. McCombie, accepted for publication in Genome Research. 18. O. Bilenko, D. Gavrilov, B. Gorbovitski, V. Gorfinkel, M. Gouzman, G. Gudkov, V. Khozikov, O. Kosobokova, N. Lifshitz, S. Luryi, A. Stepoukhovitch, M. Tcherevishnick, G. Tyshko, accepted for publication in Electrophoresis. 19. B. Ewing, L. Hiller, M.C. Wendl and P. Green, Genome Research 8 (1998) 175. 20. B. Ewing and P. Green, Genome Research 8 (1998) 186.

10