PATNEWS: PTO faces 90 year backlog on genetic patents due to sequences (fwd)

[James Love <love@tap.org>]

  ---------- Forwarded message ----------
  Date: Wed, 22 May 1996 08:56:57 -0400
  From: Gregory Aharonian <srctran@world.std.com>
  To: patent-news@world.std.com
  Subject: PATNEWS: PTO faces 90 year backlog on genetic patents due to sequences
  
  !19960522  PTO faces 90 year backlog on genetic patents due to sequences
  
      Continuing this week's theme of biotech patent news stories, a recent
  article in Science magazine reports on ongoing PTO problems with handling
  sequence comparison/prior art for gene patents (Science 5/3/96, pg 643).
  
      The problem is sequence combinatorics.  For each gene sequence the PTO
  receives, the PTO checks it against the five major sequence databases (for
  example, GENBANK) to ensure relative uniqueness (I assume no one is silly
  enough to send in a sequence that exactly matches a published sequence).
  However, given the length of gene sequences, the number of sequences nearby
  in gene distance metric space can be astronomical.
  
      The PTO uses the Smith-Waterman sequence comparison algorithm to
  identify "close" matches, using two MasPar parallel processors with 16,000
  processors.  The output of the sequence search is then evaluated by an
  examiner to check for "obviousness", often with the involvement of a senior
  examiner.  Herein lies the problem, given lots of evaluations and limited
  examiner time.
  
      According to John Doll, head of group 1800 (which handles gene patents),
  it takes about 60 to 65 hours to examine a batch of 100 sequences, for a
  cost of about $5000.  However, the application fee is only $1000, and some
  applicants, such as Incyte Pharmaceuticals (Palo Alto) and Human Genome
  Sciences (Rockville) are submitting as many as 4999 DNA sequences in a
  single patent application, which would cost the PTO over $200,001 to process,
  for a single $1000 application fee.
  
      According to the PTO, it will take the PTO about a century at a cost of
  $20,000,001 to process DNA sequences in pending patents applications for
  which it has received $100,000.
  
      Last month, the PTO held hearings on this issue, reflecting Commissioner
  Lehman's sentiments that the PTO is facing an unacceptable situation.
  Quoting the Science article, "The agency will either have to give up hope
  of acting soon on most DNA sequence claims, or grab staff from other parts
  of the agency to satisfy the genomics industry.  Because the PTO now supports
  itself entirely with customer fees, this means that chemistry and electronics
  applicants would be 'taxed' to pay for DNA patents.  Lehman thinks that
  would be unfair".
  
      Definitely unfair, since the electronics (specifically computing) patent
  applicants face a similar combinatorics problem - that of software/hardware
  prior art (which is probably in a worse state, since there are no nice
  databanks/databases like there is for gene sequences).  That is, the
  application fee is insufficient for the PTO to do even a decent prior art
  analysis.  For gene sequences, we have $1000 application fees for $5000+
  search requirements.  For computing, only about $1000 or so is needed for
  a good search, but it would first require upwards of $10,000,000 to build
  a comprehensive prior art software/hardware library.  So it seems unfair,
  and would provoke tremendous uproar, to cannablize one part of the PTO to
  gather the resources to help another part of the PTO.  At least for
  genetics though, people in that industry want to see the problem solved.
  
      The fundamental problem is that for a few "new" technology areas, the
  distribution of prior art "nearby" in some concept space for that technology
  is so dense that when weighted with a cost function for a human examiner to
  search/review, the resulting net cost is not only larger than what the PTO
  is charging, but increases faster than the PTO politically could raise fees
  and/or a bureaucracy could adapt.  You can mathematically model these
  search problem (using scientometrics), something the PTO has resisted doing.
  
      There are solutions to these problems, but their unconventional nature
  requires unconventional solutions.  It's fun watching a very conventional
  IP community continue to stumble around trying conventional solutions (such
  as the Sedimentary Patent Institute).  I wonder how bad things will get
  before the unconventional solutions are given a chance (though for gene
  sequences I think the unconventional solution is a quantum computer :-).
  
  
  Greg Aharonian
  Internet Patent News Service