Skip to main content
Premium Trial:

Request an Annual Quote

Incyte Defectors Among Founders of New Protein Structure Co., Prospect Genomics


Signifying a shift in emphasis from sequence to structure in the world of high-throughput biology, two former Incyte Genomics scientists have formed a new computational genomics company, Prospect Genomics, in Belmont, Calif.

Richard Goold, former vice president of research at Incyte, and Tod Klingler, former senior director of research bioinformatics, have assembled a team of stars from academia and industry to give scientific guidance to the new venture.

Noting the presence of brim-full protein sequence databases, a worldwide surge in new synchrotron beamlines and other instrumentation, and improvements in prediction algorithms, Goold said the “time was right” to form a company to exploit new protein structure resources.

The decision was arrived at by consensus earlier this year by the company’s ten co-founders, who in addition to Goold and Klingler include David Baker of the University of Washington, Stephen Burley of Rockefeller University and Howard Hughes Medical Institute, Ken Dill, Peter Kollman, and Irwin Kuntz of the University of California, San Francisco, William Rutter, chairman emeritus of Chiron, Andrej Sali of Rockefeller, and Daniel Santi, CEO, Kosan Biosciences.

Rockefeller Database

Centerpiece of the company is Modbase, a database of more than 200,000 protein fold and domain targets derived from the trEMBL database by company co-founder Sali using his Modeller and Modpipe programs. The company has exclusive rights to the database, which is owned by Rockefeller University. The database will be used first to identify drug targets. A typical use might be at a bioinformatics stage, when some unknown sequence – perhaps turned up in an expression study – demands structural information to make it useful. Alternatively, a data-mining approach could be employed, by comparing all members of a family, then tailoring a ligand to a binding site so it is specific only to one family member.

Because of the size of the database, Sali noted that it will be feasible to turn up reliable but previously unknown structures based on very weak sequence similarity to a known structure. “Or one could ask for all models for hypothetical proteins (without known biochemical function and family membership) which are annotated as being involved in a given disease.” Sali said his group at Rockefeller is currently following up two such examples, one implicated in the biology of Alzheimer’s disease.

A future direction for the company – and one potentially more valuable for drug discovery – will be ligand binding. This will require a new database of interactions between protein target and combinatorial libraries, which will be compiled using the Dock program authored by company co-founder Irwin Kuntz, which according to Sali is the “industry standard for simulating docking interactions.”

Another goal is to add molecular dynamic information to Modbase using the Amber program of company co-founder Peter Kollman. As knowledge of structure and its interactions deepens, this information will be critical to understand protein-protein interactions, especially those in the multi-unit complexes that mediate many of the steps of the cell’s activities. Even without molecular dynamic information, though, Goold said the top third of Modbase models are already good enough for small molecule docking.

“This is not to say they couldn’t benefit from having some molecular dynamic information done on them. Tom and I have a good idea from our background at Incyte of the good targets. We’ll take a prioritized look at the models we have from Andrej and choose those genes which are of particular interest to pharmas, and put them through molecular dynamic runs, optimize them, and get them ready for docking. Clearly this part of it is not high-throughput,” said Goold.

Another need is ab initio folding, the effort to calculate folding pathways atom-by-atom, starting from first principles of bond energy, length and angle. According to Goold, company founder David Baker’s Rosetta program (no connection to Rosetta Inpharmatics) will be brought to bear on the computationally enormous problem. “Until recently no one thought the problem was solvable,” he said, “but as David Baker proved at the last CASP competition, ab initio is now a player.”

CASP (Critical Assessment of Structure Prediction) is a biannual competition in which would-be structure predictors try out their best algorithms and modeling routines on sequences that have structures that are newly solved but not released until the time of the meeting. The Lawrence Livermore National Laboratory coordinates the events.

Far from being in competition with his former employer, Goold said he expects that Incyte, Celera, and other big players are potential customers or collaborators with the new venture. “Everyone’s looking for new ways to annotate the genes they have. Basically they are stuck with sequence homology to do the bulk of their annotation – although some do expression work as well to get additional clues as to gene function – but we bring this interesting new structural perspective.”

Goold also observed that while larger organizations are probably also starting structure groups, “there’s a lot to do, and one of the main tasks is to identify and sign up the leading talent, and we’ve already done that.”

The new company has a commitment of $8 million in first round funding from a group of venture capitalists who will be named in early October. No customers are yet signed, but Goold says possible early adopters include “companies I have known from Incyte, or known to the academic founders, who are well hooked into the industrial groups.”

—Potter Wickware

Filed under