Genentech and DeepMind are examples of start-ups that achieved major scientific advances, while parallel efforts by academic groups and by large companies on the same problems were not successful.
A comparative analysis might better show the reasons behind their achievements.
The research on DeepMind is described in the timeline, while the research on Genentech is based on published historical reconstructions, as the books by Stephen Hall (1987) and by Sally Hughes (2011), on interviews with many protagonists available from the Berkeley Library Digital Collections and on conversations with former employees, including Roberto Crea, who was one of the first five employees and even before that a key author in the papers describing the initial work done at City of Hope with Genentech support (Crea et al, 1978; Itakura et al, 1977; Hirose et al 1978; Goeddel et al, 1979).
The comparative analysis is ongoing, but the following points are emerging:
1- In both cases there was initially widespread skepticism, but the novel approaches were supported by some scientific leaders.
In the case of Genentech, a demonstration of the skepticism is the grant submitted to NIH by Riggs and Itakura, from City of Hope. It was not funded by NIH but the work was done with the support of Genentech and was a key proof-of-principle result for the company. As described in the book by Hughes (2011):
"To conduct the experiment, Riggs and Itakura needed funding. In February 1976 they submitted a grant application to the NIH entitled “Human Peptide Hormone Production in E. coli.” They asked for $400,000 for a three-year project to make somatostatin using DNA synthesis and recombinant DNA technologies. They went on to state with notable confidence, considering the uncertainties involved:
The work proposed here will lead to the production of human hormone peptides in E. coli. We think that E. coli can be used to produce human hormones more cheaply and of better quality than can be made by synthetic peptide [protein components] chemistry. The availability of inexpensive, high quality human hormones will have many clinical application[s].
That fall the NIH turned down their grant application. The reviewers decided that Riggs and Itakura could not accomplish the proposed research in the stipulated three years and labeled it “an academic exercise” without practical merit."
Riggs (2021) has stated in a review paper:
"Itakura and I then wrote and submitted to the National Institutes of Health (NIH) (in early 1976) a grant application in which we proposed to chemically synthesize the gene for somatostatin, clone it in E. coli, and assay for the production of the somatostatin polypeptide. In this grant, we also stated that if using somatostatin was successful, we would use similar technology to produce insulin. The grant was reviewed moderately well but not funded. The summary statement of the rejection noted: “In conclusion, the goals reflect extremely complex and time-consuming projects which may not be reasonably accomplished in three years . . . the only possible outcome of this work would be to confirm that these manipulations can lead to the synthesis of a human peptide in E. coli. Because of the poor choice of the biological system, this appears as an academic exercise.” In hindsight, our project on somatostatin was indeed an academic exercise, but a novel one that provided strong patents and quickly led to a flourishing new industry. "
The work was in reality completed in a few months but several novel methods which were not described in the grant applications were needed, for example the strategy to synthesize a fusion protein to protect a peptide from degradation and more efficient methods for the chemical synthesis of nucleotide oligomers (Itakura et al, Science 1987).
Several large companies contacted by Herb Boyer were initially equally skeptical. Even Michael Bishop (Nobel Medicine, 1989) stated in his Introduction to Herb Boyer's Oral History: " I recall that Herb did offer me one main chance--the opportunity to be an early investor in Genentech; I declined. It seemed a dubious scheme to me."
In the case of DeepMind, Shane Legg mentioned how they started the company at a time when deep learning methods in AI were not widely accepted.
In both cases, however, some well-known scientist supported the novel approach. For Genentech Herb Boyer from UCSF, in addition to the City of Hope team. For DeepMind Tomaso Poggio from MIT.
2- Venture capital provided initial funding. This is a type of funding that does not require consensus in the scientific community.
The leading investors were Kleiner & Perkins for Genentech and Peter Thiel for DeepMind.
A policy of gradual risk reduction was adopted, with follow-up investments in steps (rounds), depending on the results obtained.
The risk reduction strategy used for Genentech was explained in detail by Tom Perkins in his oral interview (Berkeley Library Digital Collections):
" A week or so after that they had put together the nucleus of a business proposal to do genetic engineering, Bob [Swanson] brought it to me for financing. It was very conventional, in that I would put up the money, they would hire the people, and it would be a straightforward venture. I took the view that the technical risk was so enormous. I remember asking, “Would God let you make a newform of life like this?” I was very skeptical. I said that I would agree to meet with Boyer. He came in that same week, and we sat down in our conference room for about three hours. Of course, I have a background in physics, electronics, optics, computers, lasers. Biology was never a strength for me. I really didn't know what kind of questions to ask. So I said, “Let 's just go through it, step by step. Tell me what you are going to do. What equipment you 'll need. How will you know if you have succeeded? How long will it take?” I was very impressed with Boyer. He had thought through the whole thing. He had an answer for all those questions - you'll need this equipment, these basic chemicals, and take these measurements, and on and on. I concluded that the experiment might not work, but at least they know how to do the experiment.
I still felt the risk was stupendous. The next day I got together with Swanson, and I took the view that I am willing to go along with this thing but that we have got to figure out a way to take some of the risk out of it - something instead of me giving you all of the money, then you renting the facility, buying the equipment, and hiring the people. With that approach, you'll have spent maybe a million dollars by the time you get to actually performing the experiment. Then if it doesn't work, it is all over and all that money is lost.
“Can’t we figure out some way to subcontract this experiment to different institutions each of which already had part of these capabilities?” Nobody had all of the capabilities, that was very clear. In order to give some incentive to do that, to subcontract the work, I said I would be willing to finance the thing in phases, to put up less money upfront. If this thing starts to work, then I will put up more and more money at higher and higher prices, and you and Boyer will end up owning more of the company than if we just do it the conventional way. I 'll want to own most of the company if I 'm going to take all of that conventional risk. Swanson thought that was not a ridiculous suggestion. He went back to Boyer and a few days later they had come up with three institutions that could do this work."
In the case of DeepMind multiple rounds of funding also took place, before the acquisition by Google. The initial focus on video games might also be considered a risk reduction strategy, given the previous track record of Demis Hassabis in this field.
3- A start-up originated the project, but large companies later supported their efforts, after proof-of-principle data were obtained.
Eli Lilly and then other companies for Genentech, Google for DeepMind.
4- Both start-ups generated publications, and the scientists were eventually recognized by major scientific prizes.
The companies encouraged publications, to provide recognition for the company and for the scientists, and to facilitate recruitment of the best scientists. In the case of Genentech, proprietary information was covered by patents.
5- Interdisciplinary teamwork was essential.
In both cases it was noted that this was done on a scale difficult to reach for academic groups. The motivation was the success of the company and not the career of the individual, as often the case in academic labs.
One of the early employees of Genentech, Herbert Heyneker, stated in his Oral Hystory:
" In academe, the motivation is quite different. Graduate students are there to get a PhD thesis, so they focus on their little aspect. That’s all there is to it. They don’t have to integrate into a bigger project. The postdocs are there to make a name for themselves because they want to become assistant professors, so they have to publish. Those are the most productive years. But again, the goal is very personal. “What contribution can I make to a certain understanding of whatever.” It can be very individualistic. In industry, the goals are more clearly defined, but often you need different disciplines to reach them. So, indeed, out of Genentech came articles with twelve or fifteen names on them, and it was always viewed by academe as a funny way of doing science. I found the contrary; it was a very different way of doing science, because this was a demonstration that you can accomplish a lot by working together with different disciplines."
In the case of DeepMind, it has been noted as remarkable that the AlphaFold2 paper (Jumper at al, Nature 2021) had 19 authors listed as having contributed equally.
6- Young scientists played a key role.
DeepMind was founded by postdocs.
Younger scientists enjoyed a large degree of independence in the initial activities of Genentech, especially after the company acquired its own lab facilities, while senior scientists played an advisory and strategic role. According to Hughes (2011) in the early days of Genentech "The young scientists banded together into flexible multidisciplinary teams that exhibited inexhaustible engagement, camaraderie, and a willingness to pull together to reach common ends."
REFERENCES
Berkeley Library Digital Collections
Bioscience Oral Histories
https://digicoll.lib.berkeley.edu/search?ln=en&cc=Bioscience+Oral+Histories
Berkeley Library Digital Collections
Science, Tech, & Health Oral Histories
https://digicoll.lib.berkeley.edu/search?ln=en&cc=Science%2C+Tech%2C+%26+Health+Oral+Histories
Crea, Roberto, Adam Kraszewski, Tadaaki Hirose, and Keiichi Itakura. "Chemical synthesis of genes for human insulin." Proceedings of the National Academy of Sciences 75, no. 12 (1978): 5765-5769.
Goeddel, David V., Dennis G. Kleid, Francisco Bolivar, Herbert L. Heyneker, Daniel G. Yansura, Roberto Crea, Tadaaki Hirose, Adam Kraszewski, Keiichi Itakura, and Arthur D. Riggs. "Expression in Escherichia coli of chemically synthesized genes for human insulin." Proceedings of the National Academy of Sciences 76, no. 1 (1979): 106-110.
Hall, Stephen S. "Invisible frontiers: The race to synthesize a human gene." New York: Atlantic Monthly Press, 1987.
Hirose, T., R. Crea, and K. Itakura. "Rapid synthesis of trideoxyribonucleotide blocks." Tetrahedron Letters 19, no. 28 (1978): 2449-2452.
Hughes, Sally Smith. “Genentech: the beginnings of biotech.” University of Chicago Press, 2011.
Itakura, Keiichi, Tadaaki Hirose, Roberto Crea, Arthur D. Riggs, Herbert L. Heyneker, Francisco Bolivar, and Herbert W. Boyer. "Expression in Escherichia coli of a chemically synthesized gene for the hormone somatostatin." Science 198, no. 4321 (1977): 1056-1063.
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A. and Bridgland, A., et al, 2021. Highly accurate protein structure prediction with AlphaFold. Nature, 596(7873), pp.583-589.
Riggs, Arthur D. "Making, cloning, and the expression of human insulin genes in bacteria: the path to Humulin." Endocrine Reviews 42, no. 3 (2021): 374-380.