RAPID FINITE STATE MACHINE CONTROL OF INDIVIDUAL DNA

MOLECULES IN A NANOPORE

Noah A. Wilson, Robin Abu-Shumays, Elizabeth Koch, Seico Benner and William B. Dunbar

Dynamics and Control Laboratory, Computer Engineering and Biomolecular Engineering

University of California, Santa Cruz, CA, 95064 USA

Keywords:

Nanopore, single molecule control, ﬁnite state machine.

Abstract:

This paper demonstrates feedback voltage control of individual DNA hairpin molecules captured in a

nanopore. A ﬁnite state machine is used to program voltage control logic, executed on a ﬁeld-programmable

gate array, for rapid detection and regulation of hundreds of DNA hairpins, one at a time. Prompt voltage

reduction is used for extension of the dwell time of DNA hairpins in the nanopore. Then, voltage reversal

after a preset dwell time is used for automated expulsion of molecules prior to hairpin unzipping. The demon-

strated control authority of single molecular complexes captured in the nanopore device is an integral part of

our ongoing research for direct monitoring and control of enzyme-bound biopolymers.

1 INTRODUCTION

Nanopore sequencing is based on electrophoretically

driving a singe-stranded DNA (ssDNA) or RNA

molecule through a nano-scale pore (Deamer and

Branton, 2002). The potential of this technology

is high-speed, high throughout sequential identiﬁca-

tion of all nucleotides in any single DNA or RNA

molecule. Many research groups are now exploring

and developing biological and solid-state nanopores

to achieve low-cost, high throughput nanopore-based

sequencing (Rhee and Burns, 2006), in addition to

other single molecule sensing applications (Dekker,

2007).

In the biological nanopore setup, a planar lipid bi-

layer is created across a 20 µm teﬂon aperture in a

KCl solution. A single α-hemolysin protein channel

is inserted into the planar lipid. The channel (pore)

is 15 nm in length and varies in diameter. The cis-

opening of the pore is 2.6 nm wide, opening to a 3.6

nm vestibule before narrowing to a limiting 1.5 nm

width at the beginning of the stem. The remainder of

the stem up to the trans-opening is 2 nm wide. The

vestibule is large enough for double-stranded DNA

(dsDNA) to enter, but the limiting stem is just wide

enough for ssDNA to pass through. Across the bi-

layer, AgCl electrodes are used to apply a poten-

tial that produces an ionic current through the pore.

The ﬁeld created by this voltage pulls the negatively

charged phosphate backbone of the ssDNA or RNA

through the pore, passing from the cis side to the

trans side of the pore with the trans-side voltage posi-

tive. As molecules translocate, the pore becomes par-

tially blocked by the translocating molecule, causing

an momentary drop in current. These translocation

events can be characterized by the amplitude of the

blockade current and the time the molecule spends in

the pore, deﬁned as the dwell time.

We use DNA oligomer that is 79 nucleotides to-

tal in length, with a 20 base pair hairpin (20 bphp).

The hairpin is formed by the 3’ end folding over and

annealing on itself resulting in a 20 base pair region.

The hairpin is thus the double-stranded segment, with

the single-stranded segment 35 nucleotides long (4

unpaired bases in the doubled-stranded end loop).

Upon capture of the ssDNA end, the hairpin enters

the pore vestibule and remains until the hairpin is un-

zipped. A schematic of the nanopore system and an

example 20 bphp translocation event is illustrated in

Figure 1.

Regarding the resolution limits of ionic cur-

rent measurements, homopolymers of ssDNA and

block copolymers of RNA are distinguishable

based on the measurable differences in the block-

ade current amplitude or kinetics (Akeson et al.,

1999). However, translocation rates are too fast

(up to 2 nucleotides/µsec, (Akeson et al., 1999))

to identify individual nucleotides in heterogeneous

single-stranded polymers using existing biological

nanopores (Dekker, 2007). In this paper and other

A. Wilson N., Abu-Shumays R., Koch E., Benner S. and B. Dunbar W. (2008).

RAPID FINITE STATE MACHINE CONTROL OF INDIVIDUAL DNA MOLECULES IN A NANOPORE.

In Proceedings of the First International Conference on Biomedical Electronics and Devices, pages 93-98

DOI: 10.5220/0001053000930098

 SciTePress

0 5 10 15

Time (msec)

180 mV

cis

trans

Dwell Time

(I) (II) (III)

Figure 1: Schematic of nanopore and DNA, and plot of

representative ionic current signal during a 20 bphp DNA

translocation event under 180 mV applied potential. (I) At

180 mV, KCl ions pass through the open channel result-

ing in ∼64 pA current. (II) Upon capture of the single-

stranded end of the DNA molecule into the cis opening of

the pore, the ﬂow of ions is reduced to ∼20 pA. (III) Af-

ter ∼5 msec, the voltage unzips the hairpin, causing ssDNA

to pass through the pore into the trans chamber, completing

the measured blockaded event. The duration of the event is

referred to as dwell time.

studies (Vercoutere et al., 2003; Mathe et al., 2004),

DNA with single and double stranded segments is

used to increase the dwell time of nucleotides in the

pore (0.5–10 ms, depending on applied voltage and

dsDNA segment length). Another approach is to

use DNA-binding proteins (enzymes) to increasing

the nucleotide dwell time in the pore. This is being

pursued at UCSC as part of the $1000/mammalian

genome project (Golovchenko, 2005). Under an ap-

plied voltage, the ssDNA end of enzyme-bound DNA

is captured in the nanopore, with the enzyme resid-

ing on top of the nanopore being too large to translo-

cate through it. Binding of enzymes to DNA in this

conﬁguration has been shown to increase the dwell

time of DNA in the nanopore by up to two orders

of magnitude (up to 200 msec). Recently, kinetics

of Escherichia coli exonuclease I binding to ssDNA

has been quantiﬁed using voltage ramps for nanopore-

based force spectroscopy (Hornblower et al., 2007).

The voltage ﬁeld force exerted on the ssDNA causes

it to dissociate from the enzyme after several millisec-

onds before translocating. The time-to-dissociation

in turn can be correlated to enzyme binding rate con-

stants.

In (Benner et al., 2007), the interaction of DNA

with the Klenow fragment (KF) of Escherichia coli

DNA polymerase I was explored. In the absence of

KF, capture and subsequent unzipping of 20 bphp at

constant 180 mV reveals blockades with 20 pA mean

amplitude and 4 msec median dwell time. Addition

of KF and the dNTP complementary to the DNA tem-

plate base in the KF catalytic site yielded a substan-

tial increase in blockade dwell times (110 msec me-

dian lifetime for dGTP), attributable to ternary (DNA-

KF-dGTP) complexes. Closer investigation of such

blockades revealed a two-step pattern in greater than

97% of the blockades, the ﬁrst step at 24 pA mean

amplitude, and the second (terminal) step at 20 pA

mean amplitude lasting 4ms consistent with the hair-

pin kinetics alone. It was demonstrated that the tran-

sition from step one to two resulted in dissociation of

KF from DNA ﬁrst, followed by hairpin dropping into

the pore vestibule until unzipping occurred. As a ini-

tial effort at voltage control of enzyme-bound DNA,

we demonstrated efﬁcient automated detection of in-

dividual ternary complexes (< 3 msec), based on the

characteristic 24 pA amplitude, and truncation of the

blockade time by voltage reversal after 20 ms (Benner

et al., 2007)).

This paper presents an extension of the control re-

sults presented in (Benner et al., 2007). Speciﬁcally,

we demonstrate automated detection and manipula-

tion of DNA hairpins. Rapid detection (< 2 msec) is

based on computing a ﬁltered mean amplitude of the

ionic current in real time, and monitoring the mean

relative to an amplitude range consistent with DNA

hairpin blockades (20 ± 2.8 pA). Upon detection,

two methods of voltage control are demonstrated. In

method 1, dwell time extension is achieved by prompt

voltage reduction, with the reduced voltage applied

until the hairpin unzips. A voltage for capture in-

creases the number of molecules examined, and the

reduced voltage post-capture increases the dwell time

to, in principle, facilitate sequencing. In particular,

extending the life of DNA hairpins in the pore in-

creases the time within which a terminal base iden-

tiﬁcation could be acheived using machine learning

methods (Vercoutere et al., 2003). In method 2, volt-

age reduction is applied for a preset time (10 msec)

followed by voltage reversal to expel the molecule

prior to hairpin unzipping. This demonstrates our

control authority to aggregate the dwell times of hun-

dreds of blockade events. Additionally, it comple-

ments our prior work (Benner et al., 2007), conﬁrm-

ing our ability to detect DNA-enzyme blockades and

DNA hairpin blockades. Conﬁrmation of our ability

to discern between each blockade type in real time is

part of our ongoing work. Ultimately, nanopore-based

characterization of enzyme dynamics will require di-

rect detection and control of multiple DNA confor-

mations relative to the enzyme, and direct control of

enzyme-free DNA is a prerequisite toward developing

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices

this capability.

Direct control of ssDNA in a nanopore has been

demonstrated (Bates et al., 2003), in which detec-

tion of DNA is based on monitoring the raw am-

plitude relative to a threshold level. Voltage level

changes, comparable to those employed in this paper,

were commanded to explore the zero and low volt-

age effects on ssDNA-pore interactions. In contrast

to thresholding the raw ionic current amplitude, the

windowed amplitude mean calculation we have used

here ﬁlters the current noise. Additionally, detection

depends on the mean remaining within a preset am-

plitude range (< 6 pA in spread) for multiple con-

secutive comparisons. Alternative methods for sin-

gle molecule sensing and manipulation include op-

tical tweezers and atomic force microscopy (Busta-

mante et al., 2003). For example, optical trapping

has been used to sequence DNA by attaching a pro-

cessive enzyme to a polystyrene bead (Abbondanzieri

et al., 2005), (Greenleaf and Block, 2006). At present,

greater spatial and temporal resolution of single DNA

molecule polymerization has been achieved than with

nanopores. However, these methods generally require

more preparative steps, and far fewer molecules can

be analyzed over a common time period.

2 CONTROL LOGIC SETUP

The nanopore system is setup in a 0.3 M KCl so-

lution. A patch-clamp ampliﬁer, Molecular Devices

AxoPatch 200B, regulates the applied voltage and

measures the ionic current through the channel. The

data are recorded using the Molecular Devices Digi-

data 1440A digitizer, sampled at 50 kHz and low-

pass ﬁltered at 5 kHz with a four-pole Bessel ﬁlter.

The voltage control logic is programmed using a ﬁ-

nite state machine (FSM) within LabVIEW 8 soft-

ware. The FSM logic is implemented on a ﬁeld-

programmable gate array (FPGA) hardware, National

Instruments PCI-7831R. An FPGA is a reconﬁgurable

hardware platform that permits fast measurement and

voltage reaction times (1 µsec output sample time).

An FSM is a logic construct where program execution

is broken up into a series of individual states (Gill,

1962). Each state has a command associated with it,

and transitions between states are a function of system

measurements. Measurements of the pore current are

processed and passed to the FSM as inputs. Changes

in the FSM control logic are done as necessary, then

re-compiled and re-routed to run on the FPGA. This

achieves a balance between speed and ﬂexibility, by

enabling the system to react to events on the order of a

microsecond, while also allowing for the control logic

to be reconﬁgured as necessary between experiments.

Blockade events, quantiﬁed by the blockage cur-

rent and dwell time, can be detected and monitored in

real time using the FSM/FPGA. A mean ﬁlter applied

to the incoming current signal on the FPGA removes a

large portion of the peak-to-peak noise. Speciﬁcally,

every 5.3 µsec, the FPGA samples the ionic current

and computes a windowed mean amplitude based on

the previous 0.75 ms of signal. Every 0.2 ms, the

FPGA tests if the mean is within 20±2.8 pA (17.2 to

22.8 pA range). The basis for choosing this range is

that ∼20 pA is the median amplitude for DNA 20 base

pair hairpin events at 180 mV, as shown in the exper-

imental results below. If the mean enters and remains

within this range for four consecutive tests, the FSM

logic diagnoses the blockade as a DNA hairpin event.

The nominal detection time, between DNA transloca-

tion event and diagnosis of the event, is 2.0 ms; 0.75

ms for the windowed mean to ﬁrst enter the 17.2 to

22.8 pA range, and 0.6 ms for three more conﬁrmed

tests, and 0.65 ms of delay

3 EXPERIMENTS AND RESULTS

In our ﬁrst experiment, the objective was to efﬁciently

detect individual DNA hairpin events, and increase

the blockade dwell time by lowering the applied volt-

age from 180 mV to 150 mV upon detection. This

is referred to as dwell time extension control. Next,

we sought to aggregate the extended blockade dwell

times, by expelling the DNA using voltage reversal of

-50 mV after 10 ms at 150 mV. This is referred to as

dwell time aggregation control. The motivation was

to increase the nominal hairpin dwell time, and expel

the molecule before unzipping the hairpin. A typical

20 bphp event at constant 180 mV voltage is shown

in Figures 1 and 2aI. The probability histogram of the

base 10 logarithm of dwell time (Figure 2aIII, blue)

is unimodal, with median dwell time of 2.8 ms. The

median amplitude of the event plot in Figure 2aII is

20.9 pA with an interquartile range (IQR) of 1.7 pA.

Only 6% of events are in the subset range of 13-18 pA

(2aIII, yellow). For the same experiment at constant

150 mV voltage (data not shown), the events cluster

around a median amplitude of 15 pA and 87% of 150

events are in the 13-18 pA range. Thus, under exten-

sion and aggregation control for which the voltage is

reduced to 150 mV for all detected events, a larger

Certain inefﬁciencies in FPGA signal routing into the

sampling loop caused the additional 0.65 ms of delay in the

reaction time. By bringing global signals inside the sam-

pling loop, the delay has recently been eliminated, reducing

detection time to 1.35 ms.

RAPID FINITE STATE MACHINE CONTROL OF INDIVIDUAL DNA MOLECULES IN A NANOPORE

-0.5 0 0.5 1 1.5 2 2.5 3 3.5

0.1

0.2

0.3

0.4

0.5

-0.5 0 0.5 1 1.5 2 2.5 3 3.5

0.1

0.2

0.3

0.4

0.5

-0.5 0 0.5 1 1.5 2 2.5 3 3.5

0.1

0.2

0.3

0.4

0.5

0 5 10 15 20 25 30 35

100

200

Hairpin

Detect

0 5 10 15 20 25 30 35

100

200

Reduce

Voltage

Time (msec)

0 5 10 15 20 25 30 35

200

Hairpin

Detect

Reverse

Voltage

Reduce

Voltage

100 1000

Dwell Time (msec) Base 10 Log of Dwell Time (msec)

Fraction of EventsFraction of EventsFraction of Events

(I) (II) (III)

0 0.1

Fraction of Events

Figure 2: Regulation of 20 bphp dwell time using FSM control. (I) The red current signals are low-pass ﬁltered at 5kHz, the

blue signal is a mean ﬁltered current, and the red voltage signal is the commanded voltage. Typical events and corresponding

voltage signals under a) constant 180 mV voltage, b) dwell time extension control, and c) dwell time aggregation control. (II)

Event plot of DNA events, showing average amplitude vs. dwell time for each event (point). Equation (1) (line) ﬁt to events in

bII), and amplitude histogram for events within 13-18 pA (dashed line) range in cII). (III) Probability histograms of the base

10 logarithm of dwell time for all events (blue), and for subset of events in 13-18 pA range (yellow).

percentage of blockades should have a mean ampli-

tude within the 13-18 pA range.

3.1 Dwell Time Extension (Figure 2b)

Upon diagnosis of a DNA hairpin event using the

mean ﬁltered current, the command voltage is re-

duced to 150 mV until the hairpin unzips and the

DNA translocates through the pore. Using 180 mV

for capture results in more events than 150 mV, while

reducing to 150 mV extends the life of the hairpin.

Dwell time extension is useful for terminal base-pair

sequencing by machine learning methods (Vercoutere

et al., 2003). After each translocation, the FPGA re-

sets the voltage to 180 mV. A representative event is

shown in Figure 2bI. The event plot (Figure 2bII) pat-

tern shows that events faster than the nominal diagno-

sis time of 2.0 ms are unaffected by extension control,

and events with longer dwell times converge to the

∼15 pA mean amplitude as expected. The concave

trend is also consistent with an equation for event’s

mean amplitude vs. dwell time. In particular, for an

event at 21 pA (median amplitude at 180 mV) for 2.4

, and at 15 pA (median amplitude at 150 mV) for

Step changes in voltage induce a capacitive transient,

and the transient at the end of each event is ∼0.4 ms for

changing from 150 mV to 180 mV. Thus, 2.4 ms at 21 pA is

2.0 ms of detection time and 0.4 ms of transient time. While

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices

x ms, an approximate mean amplitude

I is

I =

2.4 ∗ 21 +15∗ x

2.4 + x

. (1)

When x ≈ 24 ms, as in Figure 2bI,

I = 16 pA. Equa-

tion (1) closely matches the mean amplitude vs. dwell

time data (Fig. 2bII). Also, the fraction of events

within the subset range 13-18 pA increased to 41%,

as shown in the yellow histogram overlaid on the blue

probability histogram (Fig. 2bIII).

3.2 Dwell Time Aggregation (Figure 2c)

The objective was to aggregate the dwell times of the

extended events by applying 150 mV for 10 ms upon

diagnosis of a hairpin event, followed by voltage re-

versal of -50 mV for 5 ms. The reversal time of 5 ms

is known to be sufﬁcient to clear the DNA from the

channel, prepping the pore for the next event. Aggre-

gation control would imply a measure of control over

the distribution of the events, in addition to temporal

control of individual molecular events. A representa-

tive event is shown in Figure 2cI. As before, the event

plot (Fig. 2cII) pattern shows that events faster than

the nominal diagnosis time of 2.0 ms are unaffected

by aggregation control. Within the subset range of

13-18 pA, the median amplitude is 16 pA with 0.7 pA

IQR (amplitude histogram shown in Fig. 2cII). The 16

pA median is consistent with (1), since for x = 10.0

ms,

I = 16 pA. Also in the subset range, and the me-

dian dwell time is 12.4 ms with 0.1 ms IQR. The low

IQR indicates a high degree of control over the distri-

bution of events that extend to at least 10 ms at 150

mV. The median dwell time of 12.4 ms is commensu-

rate with 2.0 ms of detection time, 10 ms at 150 mV,

and 0.4 ms due to a transient that is included at the

end of each event resulting from voltage reversal

Summary statistics for the histograms in Figure

2III are reported in Table 1.

In (Mathe et al., 2004), the authors characterize hair-

pin unzipping at a set of constant voltages by ﬁt-

ting a curve to an unzipping probability data proﬁle.

Speciﬁcally, for an unzipping time t at voltage V , the

unzipping probability is the fraction of events with

dwell time less than t, divided by the total number of

events

. For a set of t values, the unzipping probabil-

ity data proﬁle is shown in Fig. 3 for our experiments

the 0.4 ms transient varies in amplitude, assuming 21 pA is

sufﬁcient for line ﬁtting.

The transient due to the 180 mV to 150 mV change is

included within the 10 ms waiting time under aggregation

control.

The authors formulate an alternative but equivalent def-

inition for unzipping probability.

Table 1: Summary statistics for Figure 2III.

Figure No. of Median Dwell IQR

No. Events Time (ms) (ms)

2aIII

472

2.8 4.2

2bIII

31.6 62.0

2cIII

256

12.4 0.1

Blue histogram, for events within 10 to 30 pA range.

6% (27 events) within subset 13-18 pA range.

Yellow subset histogram, for events within 13-18 pA

range.

41% of the 187 events within 10 to 30 pA range.

55% of the 466 events within 10 to 30 pA range.

at 180 mV constant, under extension control, and at

150 mV constant. As in (Mathe et al., 2004), we ﬁt

1 10 100

0.2

0.4

0.6

0.8

Unzipping Probability

UnzippingTime (ms)

Figure 3: Unzipping probability data proﬁle, deﬁned as

fraction of events with dwell time less than each unzipping

time t, and line 1− exp[t/τ

] ﬁt to proﬁle for constant volt-

ages V = 180 mV (red) and V = 150 mV (blue). Charac-

teristic unzipping time constant τ

at constant voltage V is

generated by ﬁt. Symbols: ◦ for V = 180 mV, × for exten-

sion control (transitions from 180 mV to 150 mV), and ∗

for V = 150 mV.

a line to the data, revealing a characteristic unzipping

time τ

for constant voltage V . For amplitude range

10-30 pA and dwell time range 0.3-500 ms, it is re-

vealing to compare the median dwell times with the

ﬁtted τ

constants. For V = 180 mV, the median is

2.6 ms and τ

180

= 4.2 ms. For V = 150 mV, the me-

dian is 25.2 ms and τ

150

= 39.4 ms. We observe that

data trimming has a signiﬁcant affect over the quality

of the ﬁt to the data, and consequently over the value

for τ

. In contrast, the median does not vary as much,

suggesting a sensitivity of τ

to outliers, in addition to

the ﬁtting method used. For example, for a dwell time

range of 0.3-4000 ms at V = 150 mV, the median is

RAPID FINITE STATE MACHINE CONTROL OF INDIVIDUAL DNA MOLECULES IN A NANOPORE

34.9 ms (+ 9.7) and τ

150

= 55.7 ms (+ 16.3). Careful

selection and analysis of statistical models appropri-

ate for our data (with outliers always present) is part

of our ongoing work.

4 CONCLUSIONS

We have shown that single DNA hairpin molecules

captured in a biological nanopore can be detected and

reacted to using a ﬁnite state machine implemented

on a ﬁeld-programmable gate array. The dwell time

of such translocation events can be extended to gain

more signal, which can in turn be analyzed ofﬂine

using machine learning methods to yield terminal

base-pair speciﬁc signatures. The signatures can then

be used for real-time identiﬁcation of terminal base

pairs. Additionally, the ﬁnite state machine is ca-

pable of ejecting a molecule from the pore after it

has been detected but prior to unzipping the hair-

pin. Rapid DNA hairpin detection (< 2 msec) re-

lied on a mean ﬁltered amplitude, which was required

to remain within a preset amplitude range (< 6 pA

in spread) for multiple consecutive threshold com-

parisons. The method will be tuned to differentiate

DNA-enzyme blockades from DNA alone blockades

in real time as part of our ongoing work. Ultimately,

nanopore-based characterization of enzyme dynamics

will require direct detection and control of multiple

DNA conformations relative to the enzyme, and direct

control of enzyme-free DNA is a prerequisite toward

developing this capability.

ACKNOWLEDGEMENTS

E. Koch was supported by a Summer Undergradu-

ate Research Fellowship in Information Technology,

funded by NSF under grant CCF-0552688. The work

was also supported in part by NHGRI under grant

K25 HG004035-01. We thank K. Lieberman and

M. Akeson for their help in preparing the paper.

REFERENCES

Abbondanzieri, E., Greenleaf, W., Shaevitz, J., Landick, R.,

and Block, S. (2005). Direct observation of base-pair

stepping by RNA polymerase. Nature, 438(24):460–

465.

Akeson, M., Branton, D., Kasianowicz, J., Brandin, E.,

and Deamer, D. (1999). Microsecond time-scale dis-

crimination among polycytidylic acid, polyadenylic

acid, and polyuridylic acid as homopolymers or as

segments within single RNA molecules. Biophysical

Journal, 77:3227–33.

Bates, M., Burns, M., and Meller, A. (2003). Dynamics

of DNA molecules in a membrane channel probed

by active control techniques. Biophysical Journal,

84:2366–2372.

Benner, S., Chen, R. J., Wilson, N. A., Abu-Shumays,

R., Hurt, N., Lieberman, K. R., Deamer, D. W.,

Dunbar, W. B., and Akeson, M. (2007). Sequence-

speciﬁc detection of individual DNA polymerase

complexes in real time using a nanopore. Nature

Nanotechnology. advanced online publication, DOI

10.1038/nnano.2007.344.

Bustamante, C., Bryant, Z., and Smith, S. B. (2003). Ten

years of tension: single-molecule DNA mechanics.

Nature, 421:423–427.

Deamer, D. and Branton, D. (2002). Characterization of

nucleic acids by nanopore analysis. Acc. Chem. Res.,

35:817–825.

Dekker, C. (2007). Solid-state nanopores. Nature Nan-

otechnology, 2:209–215.

Gill, A. (1962). Introduction to the theory of ﬁnite-state

machines. McGraw-Hill.

Golovchenko, J. (2005). Electronic sequencing in

nanopores. The $1000/mammalian genome project,

NHGRI R01 HG003703-03.

Greenleaf, W. J. and Block, S. M. (2006). Single-

molecule, motion-based DNA sequencing using RNA

polymerase. Science, 313:801.

Hornblower, B., Coombs, A., Whitaker, R., Kolomeisky,

A., Picone, S., Meller, A., and Akeson, M. (2007).

Single-molecule analysis of DNA-protein complexes

using nanopores. Nat Methods, 4(4):315–317.

Mathe, J., Visram, H., Viasnoff, V., Rabin, Y., and Meller,

A. (2004). Nanopore unzipping of individual dna hair-

pin molecules. Biophysical Journal, 87:3205–3212.

Rhee, M. and Burns, M. (2006). Nanopore sequencing tech-

nology: research trends and applications. Trends in

Biotechnology, 24:580–6.

Vercoutere, W., Winters-Hilt, S., DeGuzman, V., Deamer,

D., Ridino, S., Rodgers, J., Olsen, H., Marziali, A.,

and Akeson, M. (2003). Discrimination among indi-

vidual watson-crick base pairs at the termini of sin-

gle DNA hairpin molecules. Nucleic Acids Research,

31:1311–8.

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices