Transition from Academia to Industry and Back - Journal of Chemical

Jul 30, 2018 - Biography. Sereina Riniker studied chemistry at ETH Zurich and completed her Master's degree in 2008. After an internship in the resear...
0 downloads 0 Views 334KB Size
Perspective Cite This: J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

pubs.acs.org/jcim

Transition from Academia to Industry and Back Andrea Volkamer*,† and Sereina Riniker*,‡ †

In silico Toxicology, Institute of Physiology, Charité−Universitätsmedizin Berlin, Charitéplatz 1, 10117 Berlin, Germany Laboratory of Physical Chemistry, ETH Zurich, Vladimir-Prelog-Weg 2, 8093 Zurich, Switzerland

Downloaded via NEW MEXICO STATE UNIV on July 31, 2018 at 08:32:02 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.



ABSTRACT: Many doctoral students and postdoctoral fellows face at some point in their career the decision between continuing in academia or pursuing a job in industry. Both career paths come with their advantages and disadvantages as well as associated clichés. Our scientific journeys have led us from an university Ph.D. degree to an industrial postdoctoral stay and back to a young faculty position in academia. In this perspective, we share our experiences while changing perspectives. We will discuss the insights we gained through the phase as industrial postdoctoral fellows, the motivation to return and take up a young faculty position in academia, and the freedom and the burden of starting out as a principal investigator (PI). We end with our thoughts on “quo vadis” computational chemistry.

I

and being able to focus exclusively on one’s research. The freedom to publish is, however, not guaranteed with all industry postdoc positions, therefore, it is very important to check beforehand that the postdoc work is not directly connected to sensitive in-house projects. Another point to be aware of when pursuing a postdoc in industry is that one is typically the only person not working on those in-house projects, i.e. collaboration with other group members is typically hindered and collaborating with other companies or university groups as often occurs in academia requires material transfer and/or confidential disclosure agreements. Administrative and teaching duties are generally low or absent when working as a postdoc fellow at a company. While this can be an advantage and definitively is a time gain concerning research, getting involved in teaching, seminars, or student supervision helps to prepare for the independent academic career. Looking back, we both consider our industry postdoc stays as having fulfilled their promise. In addition, they have provided us with a professional network that would be more difficult to establish as a young scientist staying in academia. The majority of researchers leaving academia for industry do not return to a faculty position; however, there have always been exceptions. Research is different in academia and industry, with advantages and disadvantages on both sides. In the field of computational chemistry, one of the biggest assets of industry research is clearly the size and consistency of experimental data sets, e.g. binding affinities, crystal structures of protein−ligand complexes, physicochemical properties, and

ndependently from each other, our scientific training followed a similar path (see Biographies). A highly valuable experience early on was to leave our respective home countries during or after earning our Master’s degrees to learn new scientific techniques, explore, and get accustomed to other cultures. Inspired by these experiences, we both went back to start a Ph.D. in established research groups, in Switzerland and Germany, respectively. Choosing the Ph.D. supervisor and the respective lab wisely is probably one of the most crucial steps in a scientist’s career, as a supportive principal investigator (PI) allows one to learn from example, to grow, and to become an independent researcher. Both of us had the luck to have found great mentors with our supervisors. Another important aspect is to start early on during the doctoral studies to reach out and grow a network, to talk with people in industry and academia in order to get an idea of possible professions and career paths. Although the decision about the future career direction appears to be far away at the beginning of the doctoral studies, many students are caught by surprise when they find themselves at the end of their Ph.D. without a clear idea of the next steps. Both of us decided to leave academia after the Ph.D. for a postdoctoral (postdoc) stay in the pharmaceutical industry, i.e. Novartis, or a near-pharmaceutical startup, i.e., BioMedX Innovation Center. This allowed us to get insights into the “real world” while keeping the door open to an academic career. Some of the postdoc programs or settings at companies are competitive with postdoc stays in academia as the research in computational chemistry in industry involves method development while having access to large in-house data sets. In addition, these postdoc programs provide the opportunity to work with domain experts, to get a larger overview and to acquire new knowledge, while keeping the freedom to publish © XXXX American Chemical Society

Received: July 11, 2018

A

DOI: 10.1021/acs.jcim.8b00459 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Perspective

Journal of Chemical Information and Modeling

explore new techniques and skills, to broaden the research profile, and to reach out and establish collaborations. In addition, there is the work with students. Being part of the education system, we teach the next generation of computational chemists; we can pass on our knowledge and lessons learned and help them to grow. Especially in this early phase, one has an impact on the students’ scientific development and it is inspiring to see when they become engaged in their topics. Working with young people creates a very exciting atmosphere, especially if one allows an error culture and questions project directions to implement creative ideas. Seeing the first ones graduate gives an incredible feeling of joy, pride, and also relief. Where do we, as young PIs, see the field of structural bioinformatics and computational chemistry heading? Over the past years, a major trend, that we fully support, has taken on speed: openness. Industry and academic groups increasingly share data and techniques, which can only help to improve methods and increase the outcome. In addition, public databases such ChEMBL,1 PubChem,6 and PDB7 with their rapidly increasing number of compounds, bioactivities, and structural data are expanding the possibilities of academic research. The sizes of the data sets might still be much below those in companies, but they have increased to a level that allows method development, benchmarking, and applications that were before restricted to industry. Furthermore, opensource software such as the cheminformatics software RDKit (http://www.rdkit.org/), scripts, and tools shared in personal and commercial github repositories, freely available platforms such as Knime,8 and web applications, e.g. ProteinsPlus9 (http://proteins.plus) to support working with protein structures, are changing and facilitating the availability of computational chemistry tools for research. To assist in finding the right tool, platforms such as click2drug (www.click2drug. org) or the metadatabase OmicTools10 can be used. In conjunction with this, the community is establishing benchmarking sets in different application areas of computational chemistry to ease comparison and reproducibility. One popular approach to achieve this is through blind challenges; see, e.g., the Tox21 Challenge11 for the prediction of toxicological behavior, the DREAM challenges (http:// dreamchallenges.org/) for the prediction of multitargeting drugs, or the different SAMPL12 and D3R grand challenges13 for the prediction of solvation free energies and binding affinities. Also, teaching material has become more available, e.g. through the Teach−Discover−Treat initiative, which launched challenges to develop tutorials, such as a virtual screening workflow to identify malaria drugs.14 These are exciting developments that will shape the field of computational chemistry in the future. However, additional efforts toward more robust (standard) protocols and intensive data curation will be needed to take full advantage. Together with the growing public data set sizes and advances in computer power comes the increased interest in and applications of machine learning (ML) in chemistry. To name only a few, ML techniques have been and continue to be investigated for the prediction of physicochemical properties, binding affinities, and toxicological behavior of compounds as well as for synthesis planning, force-field parametrization, or improvement of quantum−chemical approaches. For a detailed discussion, see, e.g., recent reviews about ML or artificial intelligence in drug discovery.15−17 Despite these advances, many challenges remain when applying ML to chemistry. Data sets must have a good coverage of chemical space to ensure

toxicological and pharmacokinetic data. Better inventory and infrastructure (due to a more powerful financial situation) often allow researchers to produce results faster and in a higher amount, and standard protocols as well as experienced staff reduce the noise in the data sets substantially compared to data produced by academic research groups. In addition, there is not the selection bias toward positive results as in academia, which leads for example to a skewed ratio of active to inactive compounds in public databases assembled from scientific literature such as ChEMBL.1 A known drawback of a research position in industry, on the other hand, is the limited freedom to choose research projects, especially in fundamental research, due to time restrictions and business considerations. Here, academia can provide much more freedom, although the constraints through funding agencies have increased over the past decades, especially for young PIs. Nevertheless, the possibility to establish one’s own research directions and follow one’s interests was the major driving force for us to return to academia and take up positions as young faculty. When starting as a young PI, motivation and enthusiasm is typically high and creative ideas are abundant, but one has to convince other people (funding agencies, potential collaborators) to support them either financially or through a common research effort. It is usually the first time in a scientific career when one is fully self-dependent; one has to make the decisions that will enable the business to run or not. This freedom and autonomy is thrilling, but the responsibilities that come with it (also with respect to Ph.D. students and postdocs) can at times feel like a burden. It takes time and patience until the first grant applications are approved and collaborations are established, and one gains confidence in the chosen direction and in one’s ability to lead and support a healthy and productive research group. There is often little training in a scientific career path on how to become a good supervisor, although it is one of the keys to success. Likewise, skills in administration, human resources, and teaching are suddenly needed without sufficient exposure beforehand. As this has been noted before2 and there is already useful literature available,3−5 we will not discuss these issues in more detail here but refer interested readers to these publications. However, one aspect not discussed so often is the uncertainty and pressure of a young faculty position. Even with a tenuretrack optionand many positions come withoutevaluations by the university with or without external referees take place in a 1 to 3 year cycle. Although most universities do not disclose their quantitative criteria for the evaluations, soft performance measures include any of the following: publications, acquisition of grant money and external funding, teaching, collaborations, visibility in the scientific community, faculty work, and public outreach. Thus, output has to be produced fast, while the buildup of a productive research group takes time. This conflict imposes pressure on the young PI, which at times can become difficult to cope with. It might feel like “chasing publication credit to secure career prospects”,3 but it can help to know that others are in the same situation; thus, we advise our peers to look for informal exchange with other young PIs. In addition, we see it as highly important that one tries to keep a healthy work−life balance such that the social environment can provide support in challenging times. Of course, not everything is pressure and challenges in a young faculty position, the opportunity to run an independent lab, to choose the research direction and pursue one’s interests is an enormous empowerment. One has the autonomy to B

DOI: 10.1021/acs.jcim.8b00459 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Perspective

Journal of Chemical Information and Modeling

The authors declare no competing financial interest.

broad applicability, and they must be of sufficient quality to overcome “garbage in, garbage out”. And above all, efforts to evaluate ML models, to interpret the models and to extract chemical and physical knowledge from them, are desired to go beyond mere correlations. ML methods can be extremely powerful tools, but they must be applied with savvy and care. In computer-aided drug design, many of the approaches (with or without ML) focus still on either a 2D or a static 3D description of the involved molecules. However, the majority of intermolecular interactions are governed by a mixture of enthalpic and entropic contributions. One of the reasons for retaining a more simplistic 2D or static 3D perspective is the major challenge to estimate the correct conformational ensemble of molecules. On one hand, high-throughput crystallography will boost the number of high resolution structures. On the other hand, we see with the advances of computer power an increased use of molecular dynamics and kinetic modeling as a measure to introduce a dynamic 3D perspective in drug discovery. In the same vein, advances in cryo-EM, especially with respect to structural resolution, will render this technique a valuable tool in drug design, to allow studying proteins as well as larger complexes, protein−protein interactions, and also dynamics.18 With more data, more computer power, and novel (experimental) techniques becoming available, our understanding of the underlying biological processes will broaden. This in turn will improve our abilities to rationally design new therapeutics. Nevertheless, these advances bring new challenges, e.g. an exponential increase in the amount of data, which will require novel methods to address them. Thus, it remains essential to ask critical questions and to think outside the box to find innovative solutions. Furthermore, this can only be achieved if we continue to follow the path of openness and sharing. Only with interdisciplinary collaborations, we will be able to solve more complex problems and to move to more holistic approaches.

Biographies Sereina Riniker studied chemistry at ETH Zurich and completed her Master’s degree in 2008. After an internship in the research department of Givaudan AG and a research stay at the University of California−Berkeley, she returned in 2009 to ETH Zurich to obtain a Ph.D. in molecular dynamics simulations with Prof. Dr. W. van Gunsteren. From 2012 to 2014, she held a postdoctoral position in cheminformatics in the group of Dr. G. Landrum at the Novartis Institutes for BioMedical Research in Basel, Switzerland, and Cambridge, MA. Since June 2014, she has headed the group for computational chemistry at ETH Zurich, focusing on the development of methods and software for classical molecular dynamics simulations and cheminformatics and their application to gain insights into challenging biological and chemical questions. Andrea Volkamer received her Master’s degree in Bioinformatics from Saarland University in 2007. After a one-year research stay at Purdue University (USA) working in the group of Prof. Dr. M. Lill on flexible docking algorithms, she joined the group of Prof. Dr. M. Rarey at the ZBH, Hamburg. In 2013, she defended her Ph.D. thesis with focus on computational active site and druggability predictions. After a short ProExzellezia postdoctoral period, she joined BioMedX Innovation Center in Heidelberg as a postdoctoral researcher in the group of Dr. S. Fulle, where she has been working on tools to assist the development of selective kinase inhibitors in collaboration with Merck KGaA. Since July 2016, Andrea Volkamer has been an assistant professor at the Charité Berlin with focus on method development for structural bioinformatics and in silico toxicology predictions (funded by the German Federal Ministry of Education and Research (BMBF, grant nr: 031A262C) as part of the Berlin-Brandenburg Research Platform BB3R).





CONCLUSION After getting insights into research at academia and industry, it is clear to us that both worlds have their advantages and disadvantages. Although it is not a common career trajectory for university professors, having had the exposure to industrial research during our postdoc work has proven to be highly valuable for our current positions in academia. It has provided us with a professional network and appreciation for real world problems that would have been difficult to obtain otherwise. While funding, goals, and focus are different in industry and academia, the increased openness with respect to data, source code, and collaboration have moved the disciplines closer to each other and opens up new possibilities in computational chemistry.



REFERENCES

(1) Gaulton, A.; Hersey, A.; Nowotka, M.; Bento, A. P.; Chambers, J.; Mendez, D.; Mutowo, P.; Atkinson, F.; Bellis, L. J.; Cibrián-Uhalte, E.; Davies, M.; Dedman, N.; Karlsson, A.; Magariños, M. P.; Overington, J. P.; Papadatos, G.; Smit, I.; Leach, A. R. The ChEMBL Database in 2017. Nucleic Acids Res. 2017, 45, D945−D954. (2) Van Noorden, R. Some Hard Numbers on Science’s Leadership Problems. Nature 2018, 557, 294−296. (3) Barker, K. At the Helm: Leading Your Laboratory, 2nd ed.; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, NY, 2010. (4) Norris, D.; Dirnagl, U.; Zigmond, M. J.; Thompson-Peer, K.; Chow, T. T. Health Tips for Research Groups. Nature 2018, 557, 302−304. (5) Kolb, P.; Klappstein, V.; R, T. The 15 Things that Surprised Me Most When I Started Out as an Independent Group Leader. J. Postdoctoral Affairs 2012, 2, 30−34. (6) Wang, Y.; Bryant, S. H.; Cheng, T.; Wang, J.; Gindulyte, A.; Shoemaker, B. A.; Thiessen, P. A.; He, S.; Zhang, J. PubChem BioAssay: 2017 Update. Nucleic Acids Res. 2017, 45, D955−D963. (7) Rose, P. W.; Prlić, A.; Altunkaya, A.; Bi, C.; Bradley, A. R.; Christie, C. H.; Di Costanzo, L.; Duarte, J. M.; Dutta, S.; Feng, Z.; Kramer Green, R.; Goodsell, D. S.; Hudson, B.; Kalro, T.; Lowe, R.; Peisach, E.; Randle, C.; Rose, A. S.; Shao, C.; Tao, Y.-P.; Valasatava, Y.; Voigt, M.; Westbrook, J. D.; Woo, J.; Yang, H.; Young, J. Y.; Zardecki, C.; Berman, H. M.; Burley, S. K. The RCSB Protein Data Bank: Integrative View of Protein, Gene and 3D Structural Information. Nucleic Acids Res. 2017, 45, D271−D281. (8) Fillbrunn, A.; Dietz, C.; Pfeuffer, J.; Rahn, R.; Landrum, G. A.; Berthold, M. R. KNIME for Reproducible Cross-Domain Analysis of Life Science Data. J. Biotechnol. 2017, 261, 149−156. (9) Fährrolfes, R.; Bietz, S.; Flachsenberg, F.; Meyder, A.; Nittinger, E.; Otto, T.; Volkamer, A.; Rarey, M. ProteinsPlus: A Web Portal for

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected] (A.V.). *E-mail: [email protected] (S.R.). ORCID

Sereina Riniker: 0000-0003-1893-4031 Notes

Needless to say, this is a personal perspective. Tools, software, or datasets mentioned serve as examples and are not the result of an exhaustive literature search. C

DOI: 10.1021/acs.jcim.8b00459 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Perspective

Journal of Chemical Information and Modeling Structure Analysis of Macromolecules. Nucleic Acids Res. 2017, 45, W337−W343. (10) Henry, V. J.; Bandrowski, A. E.; Pepin, A.-S.; Gonzalez, B. J.; Desfeux, A. OMICtools: An Informative Directory for Multi-omic Data Analysis. Database 2014, 2014, bau069. (11) Huang, R.; Xia, M.; Nguyen, D.-T.; Zhao, T.; Sakamuru, S.; Zhao, J.; Shahane, S. A.; Rossoshek, A.; Simeonov, A. Tox21Challenge to Build Predictive Models of Nuclear Receptor and Stress Response Pathways as Mediated by Exposure to Environmental Chemicals and Drugs. Front. Environ. Sci. 2016, DOI: 10.3389/fenvs.2015.00085. (12) Guthrie, J. P. A Blind Challenge for Computational Solvation Free Energies: Introduction and Overview. J. Phys. Chem. B 2009, 113, 4501−4507. (13) Xu, X.; Yan, C.; Zou, X. Improving Binding Mode and Binding Affinity Predictions of Docking by Ligand-based Search of Protein Conformations: Evaluation in D3R Grand Challenge 2015. J. Comput.-Aided Mol. Des. 2017, 31, 689−699. (14) Riniker, S.; Landrum, G. A.; Montanari, F.; Villalba, S. D.; Maier, J.; Jansen, J. M.; Walters, W. P.; Shelat, A. A. Virtual-screening Workflow Tutorials and Prospective Results from the TeachDiscover-Treat Competition 2014 Against Malaria. F1000Research 2017, 6, 1136. (15) Lo, Y.-C.; Rensi, S. E.; Torng, W.; Altman, R. B. Machine Learning in Chemoinformatics and Drug Discovery. Drug Discovery Today 2018, DOI: 10.1016/j.drudis.2018.05.010. (16) Mayr, A.; Klambauer, G.; Unterthiner, T.; Steijaert, M.; Wegner, J. K.; Ceulemans, H.; Clevert, D.-A.; Hochreiter, S. LargeScale Comparison of Machine Learning Methods for Drug Target Prediction on ChEMBL. Chem. Sci. 2018, 9, 5441−5451. (17) Fleming, N. How Artificial Intelligence is Changing Drug Discovery. Nature 2018, 557, S55−S57. (18) Renaud, J.-P.; Chari, A.; Ciferri, C.; Liu, W. T.; Rémigy, H. W.; Stark, H.; Wiesmann, C. Cryo-EM in Drug Discovery: Achievements, Limitations and Prospects. Nat. Rev. Drug Discovery 2018, 17, 471− 492.

D

DOI: 10.1021/acs.jcim.8b00459 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX