Biochip arrays are known and commonly used in the high- throughput analysis of test samples, typically new chemical or biological products, to identify, for example, potential therapeutic properties. Current biochip arrays are manufactured using a micro-arraying device which deposits an array of different"probes"on a substrate using a complex multi-head positioner. The biochip array is then exposed, in use, to the test sample. Each of the probes is associated with a corresponding property (such as protein binding) and produces a detectable effect, for example a change in the wavelength of a fluorescent label which is associated with the probe, if the test sample exhibits the corresponding property. The probes may contain, for example, proteins, cells or DNA sequences.

Examples of test samples that may be analysed include: (a) proteins that give an indication of a particular disease state ("diagnostic marker") ; (b) proteins or nucleic acids that give an indication of a genetic sequence variation ("genotype"); (c) small molecules that have a desired therapeutic effect ("pharmaceuticals") ; or (d) a synthetic or a natural ligand that binds with high affinity ("affinity ligand").

One of the main problems with existing biochips is that they are expensive. This is mainly because of the expense associated with the micro-arraying device used to manufacture the biochip. An additional problem is that manufacturers of biochips usually sell biochips having a large array of different molecules (such as 1000 different molecules on each biochip). Whilst this allows the biochip to be used in a large number of different experiments, often the laboratory testing the test sample is only interested in the interaction of the test sample with a few of the probes on the biochip. This is therefore wasteful especially since many of the probes that are used are expensive. Unfortunately, this problem cannot be solved by the laboratory simply asking for a biochip array with specific probes on it, since such "bespoke"biochips are significantly more expensive to produce than the standard"off-the-shelf"biochips.

According to one aspect, the present invention provides a new way of manufacturing biochips. According to this aspect, the biochips are manufactured by depositing the probes in a random manner on the substrate. As a result, there is no need to use the complex and expensive high accuracy positioning devices used in existing micro- arraying devices. The probes may be deposited on the substrate by, for example, spraying the probes onto the substrate. In a preferred embodiment, the probes are sprayed using an aerosol nozzle, since this allows more control of the size of the droplets being sprayed onto the substrate.

Embodiments of the present invention will now be described by way of example with reference to the following Figures in which :

Figure 1 is a perspective view of a biochip manufacturing machine used to manufacture a biochip; Figure 2 is a flowchart illustrating a method in which a sample is tested with a number of probes; Figure 3a is a perspective view of a biochip manufacturing machine similar to that of Figure 1, but with the addition of a light source and a camera, for monitoring the deposition of probes on the substrate; Figure 3b is a plan view of a biochip made by the biochip manufacturing machine shown in Figure 3a; Figure 4 is a block diagram illustrating the main components of the biochip manufacturing machine shown in Figure 3a; Figure 5 is a block diagram illustrating the main components of an image processing unit forming part of Figure 4; Figure 6 is a block diagram illustrating an alternative image processing unit to that shown in Figure 4; Figure 7 is a perspective view of an alternative biochip manufacturing machine which allows probes to be deposited at specified locations on the biochip; Figure 8 is a perspective view of an apparatus for measuring the spatial light distribution from a fluorescing target area; Figure 9a shows a deposited probe droplet having an underlying scratch and shows two trajectories along which

the intensity profile of the deposited droplet is measured; Figure 9b shows the intensity profile of the deposited probe droplet in a direction containing the scratch; Figure 9c shows the intensity profile of the deposited probe droplet in a direction not containing the scratch; Figure 10 is a block diagram illustrating the main components of a signal processing unit forming part of the apparatus of Figure 8; Figure 11 is a plot illustrating how an intensity profile determining unit shown in Figure 10 determines the intensity profile of a target area versus time; Figure 12a illustrates an alternative method of optically scanning a target area; Figure 12b illustrates a further alternative method for optically scanning a target area; Figure 12c illustrates an alternative method of scanning in which the same result is achieved by dividing an image of a target area into a plurality of pixels; and Figure 13 is a flow chart showing a typical drug development process.

First Embodiment Figure 1 shows an aerosol machine 101 which may be used to deposit droplets 106 containing probe molecules onto a substrate 102 to form a biochip 103. Blank substrates 102 are placed as shown onto the aerosol machine 101 and

the droplets 106 are deposited onto the substrate 102 to form a finished biochip 103. In this embodiment, the aerosol machine 101 has three reservoirs 109a, 109b and 109c, each of which stores a solution containing a respective probe molecule. The aerosol machine 101 shown in Figure 1 has three reservoirs 109 for simplicity. In practice, the aerosol machine is likely to have tens or hundreds of resovoirs so that it can deposit tens or hundreds of different probes onto the biochip substrate.

Reservoir 109a may contain, for example, a solution of an example receptor protein of molecular weight 60kDa at a concentration of 1 micromolar (0.06g/Litre) upto lg/litre. In this example water is used as a solvent for the probe molecules although other solvents may be used depending on the probe molecule being sprayed. Chemicals that are liquid (or liquefiable by heating) may be sprayed without a solvent.

To spray droplets from the reservoir 109a onto the substrate 102, an electrically operated valve (not shown) is used to connect the reservoir 109a to an aerosol nozzle 107. The aerosol nozzle 107 has an orifice of diameter 0.05 to 0.15mm at a pressure of appropriate for the viscosity of the solution concerned, so that droplets 106 having an average diameter of 0.2mm are ejected and fall onto the substrate 102. Alternatively, an ultrasonically assisted atomisation process may be used to generate droplets, vibrating at 15-80 kHz, optionally up to over 100kHz if droplets <20 microns are required.

The feed pressure in this case is only enough to provide flow to the nozzle, pressure is not required to drive droplet formation. This process may then be repeated for the reservoirs 109b and 109c. In some circumstances, it may be necessary to flush out, for example using water,

the traces of liquid from a previous reservoir before spraying liquid from the following reservoir, to avoid mixing of the liquids in the pipe (not shown) that leads from the reservoirs 109 to the aerosol nozzle 107.

The droplets 106 land on the substrate 102 to form probe areas 108. Each probe area 108 is, initially, generally hemispherical and has dimensions determined by the volume of the droplet 106 from which it was formed and also determined by the surface tension between the solvent (if any) of the droplet and the material of the substrate.

If the solvent is allowed to evaporate then the probe areas 108 will change from hemispherical to substantially circular regions containing one or more probe molecules depending on the concentration of the probe molecules in the solvent and the size of the droplets. In this embodiment, the probe molecules have at one end thereof a binder portion that attaches to the substrate, in order to fix the probe molecule to the substrate.

When compared to conventional biochip manufacturing machines, the aerosol machine 101 has the advantages that it is simpler and less expensive (thereby enabling biochips 103 to be produced cheaply) and also allows customised biochips to be produced rapidly. For example, in one mode, only liquid from the reservoir 109a is sprayed onto the substrate 102 whilst in another mode liquid from the reservoirs 109b and 109c is sprayed onto the substrate 102.

In an alternative embodiment, the aerosol machine 101 is provided with a computer interface (not shown) so that it can receive information specifying which reservoirs are to be used when making the biochip 103. Thus the computer interface allows the aerosol machine 101 to

produce customised biochips 103 in accordance with the information received. For example, the information received may specify a subset of the reservoirs 109, that is to be used to produce a customised biochip 103.

As will now be described, customised biochips 103 may advantageously be used in an interactive manner when testing new chemical compounds against probe molecules deposited on the biochip 103. In particular, if it is desired to identify which probe from a plurality of different probes interacts with a particular test sample, it is useful to use the biochips manufactured by the aerosol machine 101. For example, if there are one hundred and twenty-eight different probes and an unknown one of them is known to react or"bind"with the test sample, then when using the aerosol machine 101 to attach one probe to each biochip 103, identification of the interacting probe would require, on average, sixty-four biochips 103. However, by splitting the testing of the probes into a number of stages and by splitting the total number of different probes into two subsets at each stage, with half of the probes in each subset, it is possible to illuminate half of the probes each time.

Figure 2 is a flowchart which illustrates this process for a set of eight different probes labelled A, B, C, D, E, F, G and H. The test sample (X) is fluorescently labelled so that if it binds to one of the probes on a biochip 103, the biochip will fluoresce allowing a determination to be made (by a fluorescence detecting machine) that the biochip 103 contains the interacting probe. As those skilled in the art will appreciate, a fluorescence detecting machine typically illuminates the biochip 103 using monochromatic light of a first frequency (typically from a laser) and measures the amount of light at a second frequency produced by the

fluorescence. If light is detected then the target sample X has reacted with at least one of the different probe molecules on the biochip 103 being tested. The laser may be operated continuously or it may be strobed and the fluorescence monitored after the laser has been turned off.

Whilst the method shown in Figure 2 shows all the alternative outcomes when testing sample X against probes A to H, the following discussion will concentrate on the path that is highlighted using bold lines.

At the start of the method at step S200, there are eight different probe molecules and it is not known which of these will bind to the target sample X. At step S205 the original set of eight probe molecules is halved to define a first subset of probe molecules A, B, C, D. At step S210, sample X is tested against the first subset of probe molecules, A to D. This test is performed by making a biochip 103 on which only probe molecules A, B, C and D are present.

Such a biochip may be manufactured using an aerosol machine similar to the aerosol machine 101, but having four or more reservoirs one for each of A, B, C and D.

Alternatively, a first aerosol machine may, for example, be used to deposit droplets 106 containing the probe molecules A and B on a biochip 103 and then a second aerosol machine may be used to deposit droplets of the probe molecules C and D onto the same biochip 103.

The target sample X is then tested against probes A to D by allowing a solution containing sample X to wash over the biochip 103 on which only the probe molecules A to D have been deposited. If sample X binds with any of the

probes A to D then the biochip 103 will fluoresce under suitable illumination.

In this example, the biochip 103 used at step S210 does not fluoresce and therefore sample X must bind with one of the other probe molecules, i. e. E, F, G or H. Control then passes to step S215. At step S215 the range of probe molecules E to H is halved to define a second subset including probes E and F. At step S220 a second test is performed to determine whether sample X binds with either of the probe molecules E and F. For this test, a second biochip 103 is prepared on which only probe molecules E and F are deposited. A further sample of X is then washed over the second biochip 103 and the biochip 103 is monitored for fluorescence. In this example, this time, the biochip 103 fluoresces and step S220 has therefore established that sample X binds to either probe E or F.

Control then passes to step S225 where the subset of probes E and F is halved to define a third subset comprising only probe E. A third biochip 103 is then manufactured on which E is the only probe molecule. This third biochip 103 is then tested at step S230 with sample X and the biochip 103 is monitored for fluorescence. In this example, there is no fluorescence at step S230 and the method terminates at step S235, having established that sample X binds to the probe molecule F. Thus steps S225 and S230 identify whether probe molecule E or probe molecule F was responsible for the fluorescence seen at step S220.

If, however, no fluorescence had been observed at step S220 then control would have passed to step S240 which halves the subset of probes G and H to define a third

subset, comprising only probe G. A third biochip 103 would then be prepared having only probe G, which would then be tested at step S245 with a further sample of X.

Fluorescence would indicate that the target sample X was bound to the probe molecule G whilst an absence of fluorescence would indicate that X would bind with probe molecule H. However, this is on the assumption that one of the original eight probes will bind with the target sample X. In some circumstances, it may be preferable to replace step S250 with further steps to confirm that fluorescence does indeed occur when sample X is tested with a biochip having only probe H. If probe H does not fluoresce after being tested with sample X, then it can be assumed that either one of the probes is no longer active or there is a defect in the monitoring apparatus.

The left-hand side of Figure 2 (which corresponds to the observation of fluorescence at step S210) is equivalent to the right-hand side shown for steps S215 onwards and therefore is not discussed in detail. However, just as probe molecule H was indicated by a lack of light at steps S210, S220 and S245, probe molecule A would be indicated by the presence of fluorescence at all three of the fluorescence tests. Thus a further test could be introduced, as was discussed above for probe molecule H, to guard against the possibility of spurious fluorescence being observed at each of the three tests. For example, a spurious fluorescence could be detected by making a new biochip containing one or more of probe molecules B, C, D, E, F, G and H and ensuring that no fluorescence was observed from that biochip.

Second Embodiment Figure 3a shows an aerosol machine 301 which is similar to the aerosol machine 101 (common parts are shown using

the same reference numbers) but also has a lamp 300 and a CCD (charge coupled device) camera 304. As will be discussed below, the lamp 302 is used to illuminate the substrate 302 of a biochip 303 and the signal from the camera 304 is used to determine the positions of the different probe areas 108 on the biochip 303.

Figure 3b schematically illustrates the form of a biochip 303 made using the aerosol machine 303 shown in Figure 3a. As can be seen from Figure 3b, the biochip 303 differs from the biochip 103 in that the biochip 303 has fiducial marks 305a and 305b which are used as position references when determining the positions of the different probe areas 108. In the example biochip 303 shown in Figure 3b, three different types of probes are sprayed onto the substrate 302. The three different types of probe areas containing the respective probes are labelled 108-1,108-2 and 108-3. As a result of the random nature of the spraying of the probes onto the substrate 302, some of the droplets carrying the probes will overlap with each other. In some cases these overlapping droplets will carry the same probe molecule (as indicated by reference numeral 310 shown in Figure 3b) and in other cases the overlapping droplets will carry different probe molecules (which is indicated by reference numeral 311 shown in Figure 3b). However, in this embodiment since the position of each of the different probe areas 108 is determined, these overlapping areas can be identified and ignored in the subsequent testing experiments.

In order that the subsequent testing apparatus (not shown) can identify the different probes on the substrate 302, each biochip 303 is also provided with a unique serial number 306 that is marked on the substrate. In

this embodiment, the serial number is provided both in human readable form 306a and in machine-readable form 306b. In this embodiment, the format of the data 306b is such that the testing apparatus can readily determine the serial number of the biochip 303 through suitable image processing.

Figure 4 is a block diagram illustrating the main parts of the aerosol machine 301 of this embodiment. As shown, the aerosol machine 301 includes a controller 401 which controls the aerosol machine 301 under control of software instructions (not shown) stored within the controller 401. The controller 401 is arranged to output appropriate control signals to a lamp driver 402 for turning the lamp 300 on and off; and control signals to aerosol valve drivers 403 which control aerosol valves (not shown) used to select one or more of the reservoirs 109 as the source for the probe molecules to be sprayed from the aerosol nozzle 107.

After droplets of probe molecules have been sprayed onto the substrate 302, image data from the CCD camera 304 is processed by an image processing unit 404 to identify the position and shape of the different probe areas 108 relative to the fiducial marks 305. Once the positions of the different probe areas 108 have been determined by the image processing unit 404, the positional data is passed to the controller 401 which outputs the data via a data interface 405, so that the positional data can be made available to the purchaser of the biochip 303. In this embodiment, the data interface 405 comprises a floppy disk drive and the positional data is stored on a floppy disk together with the serial number 306 of the corresponding biochip 303.

The aerosol machine 301 also has a user interface 406 for outputting warnings and the like to a human operator and/or for receiving control commands from the operator.

For example, if one of the reservoirs 109 is nearly empty then the controller 401 can output an alert to the operator via the user interface 406.

There are various ways in which the aerosol machine 301 can deposit the different probe molecules onto the substrate 302 and determine the position of the different probe areas 108 on the substrate 302. Some of these techniques will now be described.

In a first technique, the lamp 300 is initially turned on and the camera 304 is used to take a first image of the blank substrate 302. This allows the positions of the fiducial marks 305 to be determined and also allows the serial number 306 of the substrate 302 to be determined by suitable processing of the image data received from the camera 304.

The aerosol machine 301 is then used to deposit droplets 106 containing a first probe molecule onto the substrate 302, to form the first probe areas 108-1. This is controlled by the controller 401 outputting control signals to the appropriate aerosol valve driver 403 to open, for a predetermined time, the appropriate valve between the reservoir 109 containing the first probe molecules and the aerosol nozzle 107. The lamp 300 is then illuminated again and the camera 304 is used to take a second image of the substrate. The first image is then subtracted from this second image so that the differences between the two images are only due to the first probe areas 108-1. The image processing unit 404 then processes the difference image to determine the positions

of all of the first probe areas 108-1.

The aerosol machine 301 is then used to deposit droplets 106 containing the second probe molecules onto the substrate 302, to form the second probe areas 108-2.

Again, this is controlled by the controller 401 controlling the position of the valve connecting the reservoir 109 containing the second probe molecules to the aerosol nozzle 107. The lamp 300 and the camera 304 are then used again to obtain a third image of the substrate 302. The image processing unit then subtracts the second image from this third image and from the difference image determines the positions of the second probe areas 108-2.

A similar process is then performed for the third probe areas 108-3 to determine the locations of the third probe areas 108-3 on the substrate 302.

IMAGE PROCESSING UNIT The image processing unit 404 used for this first technique is shown in more detail in Figure 5. The signal from the camera 304 is digitised by an analogue- to-digital converter (ADC) (not shown) and the four images (labelled 502-1,502-2,502-3 and 502-4) discussed above are stored in a memory 501. As described above, the first biochip 303 image 502-1 is taken before any droplets 106 have been deposited, and from this image the positions of the fiducial marks 305a, 305b and the serial number of the biochip 303 are determined. In particular, these are determined by passing the image 502-1 through a selector 503 and an image subtraction unit 505 (for this image, no subtraction is performed) to an edge detector 507. The edge detector 507 processes the first biochip image 502-1 in a conventional manner to extract

the edges in the image. The extracted edges are then processed by an analysis unit 509 which interprets the pattern of edges to determine the serial number 306 and the locations of the fiducial marks 305 for the biochip 303. This information is then stored as biochip data 510 in a second memory 511.

After the second biochip image 502-2 has been stored in the memory 501, the selector 503 selects both the first biochip image 502-1 and the second biochip image 502-2 and passes these to the image subtraction unit 505. The image subtraction unit 505 then subtracts the first biochip image 502-1 from the second biochip image 502-2 and outputs a difference image identifying the differences between the two images. The only differences between the two images should be as a result of the deposition of the first probe areas 108-1. This difference image is then passed to the edge detector 507 which detects the edges of the first probe areas 108-1.

Finally, the analysis unit 509 determines the positions of all of the first probe areas 108-1 from the detected edges and stores these positions as probe #1 location data 512-1 in the memory 511.

The positions of the second probe areas 108-2 are determined when the third biochip image 502-3 is taken and stored in the memory 501. The selector 503 and the image subtraction unit 505 are then used to subtract the second biochip image 502-2 from the third biochip image 502-3, to generate a difference image that should only result from the second probe areas 108-2. This difference image is then processed by the edge detector 507 and the analysis unit 509 to determine the locations of the second probe areas 108-2. This position data is then stored as probe #2 location data 512-2 in the memory

511.

A similar process is carried out after the fourth biochip image 502-4 is stored in the memory 501, in order to determine the locations of the third probe areas 108-3, which position data is stored as probe #3 location data 512-3 in the memory 511.

The data stored in the memory 511 is then read by the controller 401 and is passed to the data interface 405, where it is recorded on a floppy disc for use with the biochip 303.

A second technique for depositing the different probe molecules onto the substrate 302 and for determining the position of the different probe areas 108 on the substrate 302 will now be described. In the first technique described above, the different probe molecules were deposited on the substrate sequentially and images of the biochip were taken at each stage. Alternatively, each of the reservoirs 109 may contain one of the different probes together with a different fluorescence molecule, which fluoresces at a different frequency to the fluorescence molecules provided in the other reservoirs 109. As a result, the three different types of probe areas 108 will fluoresce at a different frequency when suitably illuminated. The different locations of the different probe areas 108 can therefore be determined from a single image of the biochip 303 through suitable processing.

The image processing unit 404 used for this technique is shown in Figure 6. As shown, a single image of the biochip 303 is passed through three image filters 601-1, 601-2 and 601-3. Each of these filters is a band pass

filter centred at a respective one of the three fluorescing frequencies (F1, F2 and F3) of the three fluorescing molecules. In this way, the image output from filter 601-1 should only include the first probe areas 108-1 ; the image output from filter 601-2 should only include the second probe areas 108-2; and the image output from filter 601-3 should only include the third probe areas 108-3. The filtered images are then passed sequentially through the edge detector 507 and the probe locater 605 to determine the probe #1 location data 512- 1, the probe #2 location data 512-2 and the probe #3 location data 512-3, which are stored in the memory 511 as before. This data is then output together with the serial number and fiducial position data (not shown) to the controller 401 as before.

Third Embodiment Figure 7 shows an aerosol machine 701 which is similar to the aerosol machines 101 and 301 except that it includes a deflector 702 for electrostatically deflecting the droplets 106 as they fall down onto the biochip 303.

In this embodiment, the nozzle 107 of the aerosol machine 701 is connected to a high voltage supply (not shown) so that the droplets leaving the nozzle 107 are electrically charged.

The deflector 702 comprises a pair of deflector plates 706a, 706b to which a differential voltage is applied to deflect the droplets in an X direction and a pair of deflector plates 707a, 707b to which a differential voltage is applied to deflect the droplets in a Y direction. A deflection controller 704 applies suitable voltages to the deflector plates 706,707 under the control of the controller 401, so that the droplets land at substantially predetermined locations on the biochip

303, such as in a regular array as shown in Figure 7.

The aerosol machine 701 of this embodiment thus allows droplets containing probe molecules to be deposited at predetermined locations on the biochip 303 without requiring the use of any moving parts. The lack of moving parts allows the aerosol machine 701 to be manufactured relatively inexpensively, thereby allowing biochips 303 also to be produced inexpensively.

As those skilled in the art will appreciate, in some circumstances it may desirable to discharge the droplets before they land on the biochip 303, for example to prevent mutual electrostatic repulsion from perturbing the positions where the droplets land. This discharging of the droplets may be achieved by passing them through a region of oppositely charged ionised air. For example, positively charged droplets could be passed through a region of negatively ionised air molecules, which may be established by placing a negatively charged metallic tube (not shown) between the deflector 702 and the biochip 303.

Fourth Embodiment Figure 8 is a perspective view of a testing apparatus which can be used to test samples with the biochips manufactured in any of the above embodiments.

Conventional test apparatus 1 wash the test sample over the biochip 303 and an imaging system detects probe areas that fluoresce and determines the intensity profile of the fluorescence as it increases and decays with time.

The test apparatus then processes these intensity profiles (which are usually different for different probes) to make quantitative measures of the reaction of the test sample and the probes. The test apparatus may consider the fluorescence of each probe area individually

or it may use statistical processing techniques to combine the fluorescence profiles from several probe areas containing the same probe molecules. The measures most commonly used include the steepest gradient of the profile (s) and the peak intensity of the profile (s) relative to a noise floor. Unfortunately, with conventional testing apparatus, it is not possible to distinguish fluorescence of the test sample that has attached to random sites on the biochip (such as to scratches) and fluorescence of the test sample that has attached to one of the probes. Further, the imaging system itself is not capable of detecting fluorescent signals below a certain threshold, because of noise inherent in the detector and associated circuitry.

The test apparatus shown in Figure 8 has been designed to address these problems. As shown, the test apparatus 1 includes a base 3 and is positioned in use on a table 5 via four anti-vibration feet, three of which are visible and referenced 5a to 5c. In use, a shield (not shown) also surrounds the testing apparatus 1 to prevent vibrations caused by air currents.

A worktop 7 is mounted on an x-y direction translation stage 9, which is in turn mounted on a y-direction translation stage 11 that is mounted on the base 3. A head device having a housing 13 is mounted on a z- direction translation stage 15, which is in turn mounted on the base 3. As shown, each of the translation stages 9,11 and 15 moves along a respective bearing slide 17a, 17b, 17c having a dove-tail profile. The translation stages are driven along the bearing slides 17 by respective motorised lead screws (not shown). In this embodiment, the translation stages 9,11 and 15 are used only to provide coarse alignment between the worktop 7

and the head device, and a positioner (not shown) within the head device is used for fine position alignment. The positioner used for fine position alignment is preferably the one described in the applicant's co-pending International patent application PCT/GB02/00122, the contents of which are incorporated herein by reference.

By providing the positioner, the test apparatus 1 can move a detector (in this embodiment an optical fibre 21) rapidly and precisely over a selected probe area 108 whilst it is fluorescing. In particular, the test apparatus 1 uses the position data associated with the biochip 303 to identify the location of the selected probe area 108 and then moves the optical fibre 21 to that location. The test apparatus 1 then moves the optical fibre in a known manner in different directions across the probe area 108 in order to obtain intensity profiles identifying how the intensity of fluorescence of the probe area 108 varies spatially across different parts of the probe area. The signals obtained from the optical fibre 21 can then be processed to remove some of the above described noise components, to thereby provide more accurate quantitative measurements for the reaction taking place between the test sample and the probe.

This scanning of the optical fibre 21 across the probe area 108 is illustrated in Figure 9. In particular, Figure 9a shows the probe area 108 containing one or more probe molecules that are fluorescing due to interaction with the test sample. Initially, the positioner moves the optical fibre 21 along the line 903 from position x to position x'. It then moves the optical fibre along line 905 from position y to y'. The resulting signals 904 and 906 obtained from a photodiode (not shown) at the other end of the optical fibre 21 are shown for these two scans in Figures 9b and 9c respectively. In this

example, the underlying spatial profile of the two intensity signals 904 and 906 is a Gaussian. This is represented by the Gaussian plots 908 and 910. The intensity profiles 904 and 906 do not exactly correspond to the underlying Gaussian profile because of the noises discussed above.

Additionally, the intensity profile 904 shown in Figure 9b also includes a noise portion 907 caused by a scratch 901 on the surface of the substrate 302. This scratch has added to the intensity profile because some of the test sample molecules have attached themselves to this scratch 901. In this embodiment, the test apparatus 1 uses a signal processing unit (not shown) effectively to remove these noise components in order to obtain a better indication of the illumination intensity for the probe area 108 at the current time point. Once the test apparatus has obtained this intensity value, it performs another scan along the lines 903 and 905 in order to obtain a new intensity value for a subsequent time point.

In this way, the testing apparatus can build up a more accurate profile of how the fluorescence intensity of the probe area 108 varies with time, from which it can determine more accurately the above mentioned steepest gradient and peak intensity values. As those skilled in the art will appreciate, in order that this scanning process does not add to the overall noise of the final intensity plot, the above scanning must be performed at a fast enough rate that the fluorescence of the probe area 108 does not change between the scan in the x direction and the scan in the y direction.

The way in which the test apparatus 1 performs the signal processing described above will now be described with reference to Figures 10 and 11. Figure 10 shows the

photodiode 1001 which receives the light from the optical fibre 21 shown in Figure 8. The photodiode 108 will therefore output the intensity signal 904 shown in Figure 9b as the optical fibre 21 is scanned across the probe area 108 in the x-direction. This profile signal 904 is then passed to a comparison unit 1003 where it is compared with a number of predetermined template models 1005. In this embodiment, several template models 1005 are provided since it is expected that not all of the intensity profiles will be a single Gaussian. In particular, the inventors have noted that sometimes the profile has a dip in the centre. The inventors believe this is because of the"splashing"effect of the droplet as it is dropped onto the substrate 302.

In comparing the received signal 904 with the template models, the comparison unit 1003 identifies the template model most similar to the received signal 904. The signal 904 is then passed to a signal fitting unit 1007 where, for example, a least squares processing technique is used to fit the received signal 904 to the determined best matching template model. When the best matching template model is a Gaussian template such as those shown in Figure 9, this simply involves determining the peak and variance of the Gaussian which best matches the received signal. This information is then passed to an intensity determining unit 1009 which uses standard equations to determine a measure of the total intensity for the scan. In this embodiment, the intensity determining unit 1009 uses the calculated peak value and variance to determine the area under the corresponding Gaussian function 908 which it uses as the measure of the intensity value. This value is then stored in the memory 1011. A similar processing is then performed for the signal obtained by scanning along the line 905 shown in

Figure 9a, to obtain a second intensity measure.

The intensity measures for the x and y directions are then combined to obtain a combined intensity value for a current time point. In this embodiment, these values are combined by determining their average. The whole procedure is then repeated to obtain a similar combined intensity value for a next time point. This is illustrated in Figure 11, which shows the combined fluorescence intensity values Il, I2, I3etc. obtained at times tl, t2, t3 etc., plotted as an intensity profile.

This profile is then processed by an intensity profile processing unit to determine the above described steepest gradient and peak intensity values.

DRUG DEVELOPMENT PROCESS Figure 13 is a flow chart illustrating a typical drug development process that uses the biochips manufactured in the above way. Firstly, in step sl, a new sample is prepared and then the new sample is tested, in step s3, for potential therapeutic properties. In this embodiment, the testing for potential therapeutic properties includes the step of using a biochip made using the aerosol machine described in any of the first three embodiments or using the test machine described in the fourth embodiment.

If no potential therapeutic property is discovered, the process ends, whereas if a potentially therapeutic property is discovered, clinical drug development is performed in step s5. There are many reasons why a drug entering clinical development does not result in a commercial product. For example, the drug could have an adverse side-effect or could be prohibitively expensive to manufacture commercially. If the new samples do not

satisfy, in step s7, the requirements for commercial manufacture, then the drug development process ends. If, however, the clinical drug development is successful, then a drug is commercially manufactured in step s9.

As those skilled in the art will appreciate, an analogous development process will be performed for biological products.

Modifications And Alternative Embodiments A number of modifications and alternatives to the above embodiments will now be described.

In the third embodiment described above, the droplets from the aerosol nozzle were deflected using electrostatic deflectors. As those skilled in the art will appreciate, other types of deflectors may be used.

For example, magnetic, acoustic or optical deflectors may be used. For example, planar acoustic speakers may be provided to set up acoustic standing waves with nulls defining the desired pattern of droplets to be formed on the substrate. Alternatively, so-called optical tweezers may be used in which several laser beams are focussed and overlap with each other to define nulls which also deflect the droplets in the required way. Alternatively still, a mask may be placed over the substrate with holes in the mask corresponding to the locations on the substrate that probes are to be deposited. However, such an embodiment is not preferred because of the wastage of the probe material landing on the mask.

In the second and third embodiments described above, biochips were made in which the position of the droplets on the biochip were known and the position data was associated with the respective biochip. In those

embodiments, the position data identified the positions of all the probe areas on the biochip. In an alternative embodiment, the position data may be associated with a predetermined region of the biochip, such as the left- hand corner. The testing apparatus can then use knowledge about the positions of the different probes in that region to identify the type of probes on the rest of the biochip. This is possible, because the probe areas of the same type will all react in substantially the same way to a test sample. For example, the intensity profile determined for the probe areas in the region can then be used to identify corresponding probe areas on the remainder of the substrate, by identifying other probe areas having the same intensity profile. The identity profiles from the same type of probes can then be combined together, for example by averaging, to derive the measures associated with the reaction.

In the above embodiment, a single biochip was made by spraying droplets of probes over individual biochips.

In an alternative embodiment, a larger biochip may be made in the same way and then broken up into individual biochips which can then be reattached to other biochips in order to form a mosaic of biochips sprayed with different probes.

In the first three embodiments described above, an aerosol machine was described having a single aerosol nozzle. As those skilled in the art will appreciate, multiple nozzles may be provided either to be able to spray a larger area of biochip or to spray a respective different solution onto the substrate. For example, a respective different aerosol nozzle may be provided for each of the different reservoirs containing different probes.

In the second and third embodiments described above, droplets of probes were sprayed onto a biochip substrate and the positions of the different probes were determined using white light photography. By identifying the areas with a particular probe it is possible to add more specific samples, or to alter in some way specific binding sites (for example, using electrostatic charges to disrupt some probe sites). This therefore allows more specific biochip array management.

In the fourth embodiment described above, a detector was scanned across a selected probe site in an x direction and a y direction. As an alternative, a single scan may be used. However, preferably as many scans as possible are used as this minimises the noise errors. As a further alternative, instead of scanning across the entire probe area, the detector can be moved in a known way over the edge of the probe area, with the knowledge of the known movement being used to remove noise components from the detected signals. The known movement may be, for example, dithering the detector across the edge at a known frequency. Such an embodiment is illustrated in Figure 12a, which shows a probe area 108 and an arrow 1201 illustrating the dithering movement of the detector. The signal from the detector will vary in phase with the dithering motion. The detected signal can then be demodulated with respect to the dithering movement to give a signal that varies with the gradient of the fluorescent probe concentration in the vicinity of the site. During the process of demodulation, any noise signals that bear no correlation to the movement of the detector are removed.

In the fourth embodiment described above, a detector was scanned across a probe area to determine the spatial

intensity profile across the probe area. In an alternative embodiment, the detector may be fixed and the biochip may be moved in order to perform the scanning operation.

As a further alternative, the detector may be dithered in an epicyclic manner around the probe area (illustrated in Figure 12b), such that a mean signal can be derived that is composed of measurements of gradient at several positions, rather than at just one position on the edge of the probe area 108. In this way, errors from background noise are further reduced.

In the fourth embodiment described above, an optical fibre was used as the detector that was moved across the probe area. As an alternative, an ion selective or pH electrode may be used as the detector and moved across the probe area. As a further alternative, a CCD camera and lens may be used to provide the sampling of the intensity profile over the probe area, provided the probe area is covered by plural pixels of the CCD camera. Such an embodiment is illustrated in Figure 12c, which shows a probe area 108 and the corresponding pixelisation 1203 of the probe area 108 by the CCD camera.

As those skilled in the art will appreciate, the ability to sample the fluorescence over a probe area significantly increases the number and types of measurements that can be made using the test apparatus.

For example, the system can be used to precisely sample the concentration gradient of an agent established by diffusive processes within a single region by sampling across the region. The signal recovered can then be analysed as a function of known position within the area, deriving response of the assayed reaction as a function

of the position and hence the concentration of the diffusing species.

Instead of using a CCD camera in which the resolution of the CCD camera is such that multiple pixels correspond to the probe area, the spatial intensity profile can be determined using a CCD camera having a much lower resolution, for example, in which the CCD pixel resolution corresponds to or is slightly greater than the probe area. In this case, by causing the relative movement between the camera and the probe area, it is possible to read the signal from adjacent CCD pixels to determine the intensity profile. This is possible, provided the CCD can be read faster than the relative movement between the probe area and the CCD camera. For example, if there are two pixels (A and B) then the following signals may be obtained. Initially, when the probe area is completely within pixel A, the signal levels from the two pixels may have a value of one and zero for pixel A and pixel B respectively. Then, at the next CCD sampling time the CCD camera and/or the probe area will have moved such that 30% of the probe area will be covered by pixel A and 70% will be covered by pixel B and accordingly their output signal levels will be 0.3 and 0.7 respectively. Then, at the next CCD sampling time, 60% of the probe area will be covered by pixel A and 40% will be covered by pixel B. Consequently, the signal levels from pixel A and pixel B will be 0.6 and 0.4 respectively. With enough time steps, the above information can be used together with the expected profile of the intensity to determine the spatial intensity profile for the probe site.

In the first three embodiments described above, an atomiser nozzle was used to atomise a liquid into

droplets to spray them onto a biochip substrate. In an alternative embodiment, an inkjet head may be used to eject droplets of liquid (containing the probe) onto the substrate. The inkjet head may be of the piezo-electric type. In general, the use of an aerosol nozzle is preferred as such nozzles may be manufactured more cheaply than inkjet heads. However, inkjet heads have the desirable property that they eject droplets which have a relatively consistent volume, thus allowing the probe areas to also have a consistent size. Although inkjet heads are capable of depositing droplets onto a substrate with a high degree of positional accuracy, this accuracy is not needed in the present invention.

Therefore, what would otherwise be"reject"inkjet heads (i. e. those that do not eject droplets with a sufficiently uniform trajectory for them to be used in printers) may be used. It is envisaged that a plurality of inkjet heads, each having its own reservoir, could be provided on a carousel thereby allowing rapid selection and deposition of probes in accordance with received data. For example, when using the method of Figure 2 for analysing the interaction between a target X and a set of probes, the received data would specify which inkjet heads and reservoirs of the carousel should be used.

When droplets containing probe molecules are to be deposited onto a substrate it is known to provide the probe molecules with a portion that will attach to the substrate. Having the probe molecules attached to the substrate allows the finished biochip to be rinsed in a solution of a target X without washing the probes off the surface of the substrate. However, it is not essential that the probes are attached to the surface of a substrate. For example, droplets containing the target X could be placed on the probe sites directly using a

manipulator head or atomised and sprayed onto the biochip. In this case the number and volume of target droplets sprayed onto the biochip should be sufficiently high so as to ensure a high probability of one or more target droplets landing on a probe area whilst not so high that the droplets coalesce (to avoid them moving about, or running off, the surface of the biochip).

In the embodiments described above, fluorescence was used as a way of identifying when a target became bound to a probe. As those skilled in the art will appreciate, any suitable method may be used for identifying interactions between targets and probes. For example, radioactive targets could be used and the distribution of radioactivity on a biochip measured.

In the second and third embodiments described above, the biochip was provided with a serial number in both human readable form and in a machine readable form. The machine readable form was implemented as a two- dimensional barcode on the surface of the substrate. In alternative embodiments the biochip substrate may be provided with a magnetic strip on which the'serial number of the biochip may be recorded. The serial number may either be prerecorded onto a blank substrate and read by the aerosol machine or the aerosol machine may record a serial number onto the magnetic strip during the manufacture of the biochip. By providing the magnetic strip with a sufficient data capacity, the identities, and locations of the biochip, of the deposited probes may be recorded onto the biochip itself. As an alternative to a magnetic strip, a radio frequency (RF) tag may be used to receive and store data on the biochip. An RF tag uses an integrated circuit and a non-contact mechanism to read and write data from/to the RF tag. Typically,

an alternating magnetic field is used to both transmit data to the RF tag and to provide a power source for the RF tag; the RF tag transmits data by modulating the alternating magnetic field.

The substrate from which a biochip is manufactured may be glass, ceramic, plastic, metal or a living tissue.