Thursday, 7 January 2010

T 18/09 – A Pipeline Ain’t No Motorway

We have already studied several passages of this decision. When discussing the inventive step involved in the claims of the request on file, the opponent pointed out that the claimed subject-matter would have been obtained by means of “pipeline screening”. The Board dismisses this approach as being unrealistic.

[…] The [opponent] proposed also “pipeline screening” as a way which would be envisaged by the skilled person to arrive at the claimed subject-matter, i.e. a method based essentially on an analysis of freely available, unidentified EST (Expressed Sequence Tag) sequences for similarity to sequences available from a publicly available database of annotated protein sequences using publicly available search software. [43]

Whereas an automated pipeline screening might avoid the selection of a particular EST clone with the hindsight knowledge of the patent-in-suit, the screening of the complete EST sequences of a public EST database (such as those of the IMAGE Consortium) against the sequences of a public protein database (such as Swiss-Prot) would not have been performed out of idle curiosity, in view of the sheer size of these databases and the effort and resources required, but only with a specific technical purpose or target in mind. Indeed, depending on the intended purpose, the search programs, values and parameters used to carry out the automated pipeline screening would have to be defined and selected accordingly by the skilled person. In fact, there is a difference between looking for members of a protein family which are known to have high sequence identity and looking for members of a protein family having only a low sequence identity. Again, differences might arise when sequence identity is known to be homogeneously spread across the complete sequence of the proteins or when it is only limited to a unique (highly) conserved domain. [44]

Evidence for such a selection is found in document D86 which refers to the use of both BLAST and FASTA, the latter used for searching copies of specific (spliced-leader) sequences. Evidence is also on file showing that FASTX and BLASTX were also available to the skilled person at the filing date of the patent-in-suit and that the results obtained might be different depending on the search program used and on the parameters and values selected for running those programs. [45]

In the view of the [opponent], document D86 shows that automated pipeline screening of EST clones of a cDNA library was a standard method at the filing date of the patent-in-suit performed by the skilled person with no purpose in mind other than the broad determination of putative encoded proteins. It is noted, however, that the method disclosed in document D86 does not use all the EST clones of the cDNA library described in that document but, as a first step, refers to a random selection of EST clones that significantly reduces the number of distinct clones and EST sequences to analyze (585 clones generating 720 EST sequences). That number is much lower by far than that of the EST sequences present in public EST databases (GenBank, EMBL, IMAGE Consortium) available to the skilled person at the filing date of the patent-in-suit. The sheer number of EST sequences in these public databases (hundreds of thousands) renders any comparison of time, effort and resources completely inappropriate. [46]

Indeed, not even the [opponent] appears to rely on such a broad approach. By pointing to the references in document D86 to the searches of daily updates of these public EST databases and to the pre-selection of sequences to remove high copy number cDNAs, as well as to other possible criteria (such as the removal of EST sequences encoding proteins of known and/or highly present protein families), the [opponent] acknowledges that an appropriate starting point for an automated pipeline screening would not be the complete EST sequences of a public EST database but a particular selection of these sequences. [47]

In view of the large number and particular nature of possible criteria that may be used to carry out the selection of those EST sequences as well as the programs, parameters and values that should have to be selected accordingly, the board comes to the conclusion that no particular prior art, i.e. no particular group of EST sequences, has been clearly and unambiguously identified as a starting point for an automated pipeline screening nor has a document been clearly identified as representing that closest prior art. Moreover, there is no evidence on file demonstrating with certainty that each and every one of the suggested selections, or at least most of them, would have contained at least one EST sequence encoding (part of) the Neutrokine-α sequence of the patent-in-suit, such as those shown in documents D2, D22 or D24, and that with an automated pipeline screening the skilled person would have been in a position to recognize that sequence immediately and successfully retrieve it. Thus, in the board’s judgment, the pipeline approach is untenable. [48] 

I feel I have said enough on biotech, for the time being. 

To read the whole decision, click here.