Computer techniques, ranging from basic data warehousing through to de novo molecular design, make a substantial contribution to new drug research. These techniques are the subject of substantial investment and research, but how can this investment be protected?
Patents are the best-known form of intellectual property, and by and large it is true that they are the strongest and therefore most effective. In practice, the effectiveness of patents is limited in two ways: what can in fact be patented, and how well will the granted patent stand up to attack? Although in theory these come down to the same legal question - what is patentable? - they are assessed by different tribunals which interpret and apply the question completely differently. Worse, even in territories such as Europe, where in theory the tribunals are applying the same law, divergent interpretations can lead to diametrically opposite results.
This unfortunate situation applies whatever the field of technology, but in the field of computer-implemented inventions (as many bioinformatics techniques tend to be) the underlying law is also unclear. The European Patent Convention sets out the primary requirements for patentability of inventions in Article 52(1) which provides that: “European patents shall be granted for inventions which are susceptible of industrial application, which are new and which involve an inventive step.” Unfortunately Article 52 also attempts to define what things are not patentable. The provisions which concern software inventions are:
“(2) The following in particular shall not be regarded as inventions...:
(a) discoveries, scientific theories, mathematical methods;
(b) aesthetic creations;
(c) schemes, rules and methods for performing mental acts, playing games or doing business and programs for computers;
(d) presentations of information.
(3) The provisions of paragraph 2 shall exclude patentability of the subject-matter or activities referred to in that provision only to the extent to which a European patent application or European patent relates to such subject-matter or activities as such.” The patenting of data analysis methods and software related inventions has faced problems in Europe because of these provisions.
The exclusion of theories, mathematical methods, and so on is founded on the objection that these are too abstract to be covered by the system. First, a scientific theory may have no practical, industrial application whatsoever - for an extreme example, consider the theory of general relativity and the possibility of bending light beams in a gravitational field such as that produced by a body the size of the sun. Secondly, a fundamental discovery or mathematical method as such, may have potential applications in an extremely wide range of possible fields so that to grant a 20-year monopoly to the discoverer or deviser could operate to hinder technological progress disproportionately. The exclusion is limited to the principle: it is perfectly acceptable to patent a useful application of a theory or method even if the application is obvious once the discovery has been made. 
The difficulty comes in attempting to combine methods such as those specified in Article 52(c) with computer technology. Where a method is carried out by a computer, the observer might find the entire operation an excellent illustration of the practical usefulness of technology, but legally there may simply be an implementation by a known computer of an excluded category. For example, the exclusion of methods of performing mental acts has rarely appeared in the EPO case law but was relied upon by the UK courts in Raytheon Co’s application  and Fujitsu Ltd’s application.  These related, respectively, to a method of identifying unknown ships, which involved digitizing an image of the unknown ship, comparing the resulting image with a stored database of silhouettes and by choosing the closest silhouette, identifying the ship, and a method of modeling chemical molecules, In each case the court concluded that the invention claimed was a mental process and carrying it out by using a computer did not make it less so. The program merely permitted a conventional piece of apparatus to carry out an excluded act.
The EPO Appeal Board has recently invoked this exclusion in rejecting an application for a patent for a machine translation program. This algorithm is a good example of the sort of skilled work of drawing parallels between signifiers in two databases (vocabulary) and applying flexible rules (syntax) in order to re-express the meaning of the original in the terms of the second. Although the skills involved are qualitatively different from those involved in comparing the properties of molecules represented by data (for example, atomic co-ordinates, molecular weights, clogP and hash codes), there is a degree of similarity: in each case, the machine performs an operation which could otherwise be performed by the human mind.
Labor saving patents
In such a case, the advantages conferred by the use of a computer are the avoidance of labor and error in carrying out a process. The Court in Fujitsu's application pointed out that this is nothing new or inventive: 'avoiding labor and error is the sort of advantage that is provided by computers in any event'.
In many cases it will be possible to persuade the patent office that the advantages of the software-implemented invention go beyond this. For example, the number of elements being compared from diverse sources and by diverse forms of comparison may well be beyond that which even the most skilled chemist could realistically perform. However, patents over screening methods which at base are no more than a sophisticated exercise in comparison will always be vulnerable to this form of attack, either in the prosecution phase or in front of a national court after grant.
Similarly fragile are patents over computer models of molecules. Although the model may be one which has evident value as a tool for drug design, the 3D model structure described in terms of its atomic co-ordinates, may not be patentable for several reasons. First, the co-ordinates are simply information about the molecule, and therefore are a non-technical feature of the claim. Non-technical features cannot add novelty to an otherwise non-novel claim. Second, although the model can certainly be used as a screening tool, it does not add anything over and above the physical molecule which it represents in data-space. Attempting to claim it by presenting it as the output of a model design process  will not assist since the same process may produce several structures so the claim lacks clarity and fails under Article 84 EPC. Finally, the EPO has commented that, legalities aside, it simply is not possible for them to carry out a meaningful search for relevant prior art against such a claim.
On the other hand, a new method for scoring the fit between a computer model of a molecule and the binding site on the target to which it is being presented may be intrinsically novel, technical and inventive. If so, implementing the method by using a computer is legitimately patentable. Likewise, a method for designing new ligands either by stochastic manipulation of fragments or using a structured approach based on known chemistry may be patentable in itself even if all possible outputs cannot be claimed at the start; instead, much stronger claims may be filed to particular molecules designed by the software and successfully synthesized in vitro.
In fact, the limit on reach-through claims (to molecules developed as a result of using a design or screening method) is not as great a barrier as inventors often believe. In practice, a patent over a screening method (which is not eligible for a Supplementary Protection Certificate) is almost certain to have expired by the time a marketable drug has been developed by using the method tested and approved for sale.
An alternative, and in principle perpetually renewable, form of protection for the output of a screening or analytical program is database right. This was discussed in September's Patent Insight. One issue which has been the subject of referrals to the European Court of Justice by several national courts is the question of the investment required in order to protect a database: does the purpose of the investment have to be the creation of the database, or is a database which is in effect a spin-off from some investment with some other purpose also protected? This problem is less likely to arise in the drug discovery context since databases - of molecules, binding sites, affinities and so on - are clearly the objective of the investment in both laboratory and computer-implemented work. Database right, for European companies at least, will therefore provide a valuable back up to other forms of intellectual property protection.
The greatest limitation to the value of either bioinformatic patent or database rights is the potential difficulty of proving infringement. Data processing and manipulation are processes which are unlikely to be observable by looking at the results - an identical new and effective molecule may have been designed by a competitor by a completely different route. Unless there is some source of information about what went on in the research laboratories several years prior to the first round of reported trials, only those with pockets deep enough to afford the uncertainties of US litigation with its accompanying unlimited discovery are likely to be in a position to prove whether infringement has taken place.
The value of patents over screening or data mining methods is further limited by existence of the option, for would-be infringers, of offshore preparation and use of the method. Importation of the resulting data is almost certainly not an infringement: although Article 64(2) EPC states that if the subject matter of the European patent is a process, the protection conferred by the patent shall extend to the products directly obtained by such process, information generated through using a process does not fall within this. A claim based upon the use of information alone was struck out in the English case of Kirin-Amgen versus TKT & Anr. The EPO's view is that 'process' here refers to a manufacturing process only and although there is remarkably little case law on this point,  it is likely that the national courts called upon to enforce such a patent would agree. Nor does 35 USC 271(g)  cover data produced: under US law “a patent is not a hunting license. It is not a reward for the search but compensation for its successful conclusion”.  As a result, a would-be user of either a bioinformatics tool or the data produced through using it can, if the investment is warranted (for example by the size of the licence fee demanded), avoid all issues of infringement by simply establishing their laboratory or analytical department in a country where a patent has not been obtained.
First published in the November 2002 issue of Current Drug Discovery.
 Genetech Inc’s Patent CA  RPC 147
  RPC 427
  RPC 608
 International Business Machine’s application (BL 0/399/01)
 ‘A model of the ligand binding domain of receptor X designed in accordance with the modeling method of the preceding claim’
 A case was begun on the question by Roche against Housey Pharmaceuticals in Switzerland December 2001; no decision is yet available.
 Bayer AG versus Housey Pharmaceuticals Inc 169 F Supp 2d 328, 330 62 USPQ2d 1051, 1053 (D Del. 2001)
 Brenner versus Manson 383 US 519, 536 148 USPQ 689, 696 (1966)