Transcript Elements of Pattern Recognition Theory in the Analysis of

Elements of Pattern
Recognition Theory in the
Analysis of Biosensor Signals
Reshetilov A. N., Lobanov A.V., *Ponamoreva O. N.,
Reshetilova T. A., *Alferov V. A.
G. K. Skryabin Institute of Biochemistry and Physiology of
Microorganisms RAS
*Tula State University, Tula, Russia
[email protected]
PROBE
Biosensor – scheme.
Transducers:
B
I
O
M
A
T
E
R
I
A
L
Attention on TRANSDUCER and BIOMATERIALS
T
R
A
N
S
D
U
C
E
R
SIGNAL
REGISTRATION
RESULTS
DATA
PROCESSING
electrochemical detectors like Clark type electrodes, Screenprinted electrodes, pH-sensitive field-effect transistors – small pH electrodes
Basic trends of research at our laboratory
Transducer
Biomaterial
Analyzed
compound
Purpose
Amperometric
electrodes
Clark type
Microorganisms
of the genera
Nitrobacter,
Gluconobacter,
Comamonas,
Pseudomonas,
Rhodococcus.
Yeast cells of the
genera Arxula,
Pichia.
Carbohydrates,
Environmental
monitoring
Monitoring of
fermentation
processes
Clark type
electrodes
Enzymes
xenobiotics,
alcohols
Ethyl alcohol
Monitoring of
fermentation
processes
Basic trends of research at our laboratory
Transducer
Biomaterial
Analyzed
compound
Purpose
pH-sensitive
field-effect
transistors
(FET)
Microbial cells
Sugars
Monitoring of
fermentation
processes
pH-sensitive
FETs
Enzymes
Cholinesterase
Detection of
cholinesterase
inhibitors
pH-sensitive
FETs
Immunosensors,
enzymes – labels
Horse-radish
peroxidase
Urease
Glucose
oxidase
Pesticides
Microbial cells
From: K.Riedel et al., 1997, Antonie Van Leeuwenhoek, V. 71.
The scheme of measurement of cell respiration - the intensity of
oxidation process. A cell is fixed on electrode surface.
Oxygen consumption is insignificant in the absence of substrate
and increases when substrate is added.
Electrode current is proportional to oxidation rate.
The presented picture is called a substrate portrait.
1 - ethanol
2 - glucose
3 - xylose
4 - xylitol
5 - arabinose
6 - arabitol
7 - glycerol
8 – pyruvate
9- citrate
G . oxydans № 6 /Shchelkovo/ (0.879)
G. oxydans subsp. melanogenes (0.003)
G. oxydans subsp. suboxydans (0.02)
Gluconobacter sp. B-1284 (0.021)
Gluconobacter sp. B-1285 (0.024)
G. oxydans subsp. industrius (0.15)
G. cerinus (0.101)
300
Response, %
250
200
150
100
50
0
1
2
3
4
5
6
7
8
9
Substrate, 10 mM
7 strains. High difference in sensitivity.
Sensor signals corresponding to GLUCOSE are taken as 100%.
Clark –type electrode registration.
The substrate (metabolic) portrait
G . oxydans № 6 /Shchelkovo/ (0.879)
G. oxydans subsp. melanogenes (0.003)
G. oxydans subsp. suboxydans (0.02)
Gluconobacter sp. B-1284 (0.021)
Gluconobacter sp. B-1285 (0.024)
G. oxydans subsp. industrius (0.15)
G. cerinus (0.101)
300
250
Response, %
With many substrates,
one can obtain a
unique characteristic
describing the given
type of cells – that is
“metabolic fingerprint”
(the substrate
(metabolic) portrait).
200
150
100
50
0
1
2
3
4
5
6
Substrate, 10 mM
7
8
9
Suppose there is an objective to detect three substances, which are
simultaneously present in a sample. Let us consider the case when
the substance is either present or absent in a sample.
The substances are ethanol, glucose, and xylose.
This objective is important for the monitoring of some
biotechnological processes (for example – alcohol production):
xylose (sugar) and glucose (sugar) are substrates and ethanol is a
product.
The questions are:
(1) Is it possible to use biosensors for this purpose?
(2) Is it possible to solve this problem using the elements of cluster
analysis?
The plan of solution of the above problem – a scheme
The measuring system was represented by three amperometric microbial sensors
with immobilized cells of Gluconobacter oxydans (no 6), Hansenula polymorpha
(no 3), and Escherichia coli (no 8).
10 strains were tested altogether; 3 strains were selected.
Gluconobacter
oxydans
Sensor response
Glucose Ethanol Xylose
+
+
+
Hansenula
polymorpha
-
+
-
Escherichia coli
+
-
+
Michaelis constants – biochemical
assessment
Michaelis constants
Km, mM
(G. oxydans)
Km, mM
(E. coli)
Km, mM
(H. polymorpha)
Substrate
Glucose
4,3
0,6
Sensor insensitive to
this type of
substrate
Ethanol
1,9
----
3,7
3,4
Sensor insensitive to
this type of
substrate
Xylose
43,6
Gluconobacter oxydans, Hansenula polymorpha и Escherichia coli.
An appearance of the 4-channel set-up
in the process of an experiment
Types of clusters – analyzed substances and their
combinations made 8 types of clusters
1 - Glucose, 1 mM
2 – Xylose, 1 mM
3 – Ethanol, 1 mM
4 - Glucose + xylose, 1 + 1 mM
5 - Glucose + ethanol, 1+ 1 mM
6 - Ethanol + glucose, 1 + 1 mM
7 - Glucose + xylose + ethanol, 1 + 1 + 1 mM
8 - Glucose = xylose = ethanol = 0 mM
Normalization of sensor signals
–
for the possibility of comparing the signals from different sensors in response to different
substances
S
a , norm
i

(S
S
a
i
)  (S )  (S )
a 2
i
b 2
i
c 2
i
Where Sai is the response of sensor a to sample i. The responses of
sensors S1, S2, and S3 make it possible to obtain a single vector S i in
the space of “Sensor response 1” – “Sensor response 2” – “Sensor
response 3”.

The space of clusters – glucose-xyloseethanol.
The clusters are located on the surface of a single sphere.
Sensor 2
Response
Cluster
"glucose"
Cluster
"xylose"
S2
Cluster
"ethanol"
S
S1
Sensor 3
Response
Xylose (1-3)
S3
Glucose (1-3)
Ethanol (1-3)
Sensor 1
Response
Identification of a sample by microbial sensors
– the scheme
For measurement, sensor readings are normalized in
accordance with the given formula. The formula allows
obtaining of sensor responses S1, S2, and S3, the sum of which
gives vector S. i = 1, 2, 3, 4, 5, 6, 7, 8 clusters.
39 samples were used for clusters formation – 7 centers
Then, the distance from the point specified by vector Si to the
centers of clusters is calculated.
The sample is considered to correspond to the closest cluster
(the Figure shows an example of a sample belonging to the
“ethanol” cluster).
The sample is considered to correspond to the closest
cluster (the Figure shows an example of a sample
belonging to the “ethanol” cluster).
Sensor 2
Response
Cluster
"glucose"
Cluster
"xylose"
S2
Cluster
"ethanol"
S
S1
Sensor 3
Response
Xylose (1-3)
S3
Glucose (1-3)
Ethanol (1-3)
Sensor 1
Response
The example of recognition of two compositions – distance to the
clusters is presented.
(A) glucose + xylose + ethanol and
(B) ethanol + xylose = ethanol
The sample containing glucose (gl)+ xylose (xl) + ethanol (et)
0,8
a
б
0,6
0,4
А)
0,2
0
гл+кс+эт
гл
эт
кс
гл+эт
гл+кс
эт+кс
The sample containing ethanol and xylose (et+xl)
0,8
0,6
0,4
0,2
0
гл+кс+эт
гл
эт
кс
гл+эт
гл+кс
эт+кс
B)
Task
(1) Is it possible to use biosensors for recognition?
(2) Is it possible to solve this problem using the elements of cluster analysis, i.e.
pattern recognition theory?
---------------------------------------------------------------------------------------------------------
Conclusion
It has been shown possible to identify the
components of a mixture containing glucose,
xylose, and ethanol by sensors based on
microbial cells.
Conclusion
At the analysis of 39 control samples, 37
samples were recognized correctly and
two samples were recognized falsely (the
ethanol + xylose mixture was identified
twice as ethanol only).
What was new ?
Difficult type of analysis – LC ?
Microbial cells – not enzymes (GOD, AO)
Conclusion
The possibility of combining the efforts of
specialists
• in chemometrics (pattern recognition) and
• developers of biosensors
for joint solution of the problems of detection
quality improvement is welcome. For example –
detection of false strong drinks.
Laboratory of biosensors
Thank you!
Reshetilov Anatoly N.
E-mail:
[email protected]