Combinatorial anti-cancer treatment, model-based prediction to guide preclinical testing

Development of efficient combinatorial anti-cancer treatment

Combined anti-cancer drugs may together target multiple robustness or weakness features of specific cancer subgroups or individual tumours[1], and their effectiveness can be further enhanced by exploiting synergistic drug actions that inhibit cancer growth and evade tumour resistance mechanisms more efficiently than drugs administered individually. Synergistic combinations also allow for a significant reduction in the dosage of individual drugs while retaining the desired effect, and thereby can ensure treatment efficacy without pushing single drug dosage to levels inducing adverse effects.

Modelling to guide preclinical pipelines for identification of combinatorial treatments

Originally spearheaded by S. Kauffmann[2] and R. Thomas[3], logical (Boolean) multiscale models have been shown to accurately describe molecular mechanisms underlying cellular decision making[4]. Boolean models can be hand-built from logical statements that are derived from pathway databases, general knowledge bases and the scientific literature. The correctness of models can be checked by observing that logical rules during a simulation/model updating scheme govern model state transitions that evolve to a stable state (or cycle), representing a clearly definable biological or cellular state. Many Boolean modelling software tools are available for this, and some have been used by us, in particular the software suite GINsim[5]  which we have used to build an extensive logical multiscale model for the cancer cell line AGS10. Starting with a regulatory network valid for a variety of cells and conditions, the model was configured with baseline phenotypic biomarkers from actively growing AGS cells to obtain a ‘committed model’ accurately depicting the regulatory logics of AGS cells and connecting it to the phenotype scale. After model reduction done to improve the computational tractability, batch-wise simulations emulating a combinatorial drug perturbation strategy predicted five synergistic drug combinations from a total of 21. Experimental testing of all drug combinations for their effect on AGS cell growth confirmed four of the five combinations synergistically reducing cell growth, indicating a false positive rate of only 20%. Importantly, the predictions did not suffer from false negatives indicating the efficacy of this approach to eliminate non-effective combinations without preventing potential blockbuster drug combinations from being tested. This is a key requirement for any in silico screening strategy. Our approach is therefore relevant to preclinical discovery of efficient anti-cancer drug combinations, and thus for the development of strategies to tailor treatment to individual cancer patients.

Knowledge for models depicting regulatory networks guiding cancer cell decision

Background knowledge sources, pathways, regulatory networks, causal reasoning:

The amount of biology that can be explained without the use of models is shrinking[6], Automated model assembly seems straightforward, as many knowledge bases in the public domain (e.g. Reactome (http://www.reactome.org/), Pathway Commons (www.pathwaycommons.org/), Signor (http://signor.uniroma2.it/)) contain detailed information about biological networks and pathways, and their components and relationships. For logical modelling, however, the only essential relationships are those that indicate causality: relationships between network nodes (proteins, RNAs, genes) that carry information about regulation – molecular actions that activate or inactivate another molecular component in a network. These causality statements can be obtained from pathway resources mentioned above, or from the literature, assisted through text mining efforts. Although many efforts are ongoing to accumulate and curate these types of knowledge, additional efforts are needed, for instance on the curation of DNA binding transcription factors and on signal transduction pathway components. Central to this effort will be a CNIO/BSC-NTNU-collaborative research on text-mining based information retrieval of causal statements from literature[7] building on the strong competence in literature information retrieval of the Valencia group[8].

Biomarker data: Genotype and phenotype information on cancer cell lines, patient-derived xenografts, patient data from solid and liquid biopsies and other tumour material

Thus, phenotype information (e.g. transcriptomic or cell signaling status) will be necessary to provide additional input for the development and deployment of combinatorial anti-cancer treatment[5] and for the foreseen necessary movement from druggable targets to druggable (sub)networks[27]. Indeed, the accuracy of predictive modeling of drug combination effects in pre-clinical cancer cell line models was crucially dependent on phenotype (transcriptomic) data[9].

The wealth of publicly available data for gaining insight into regulatory networks that drive cancer includes cancer omics available through TCGA- and ICGC-portals which provide genomics (genome-/exome sequence, copy number aberrations), epigenomics (mainly DNA methylation), transcriptomics and proteomics (mainly RPPA, reverse phase protein array); cancer cell line drug responses available from e.g. the CCLE project;  and data on genetic vulnerabilities affecting cancer cell line viability, determined by genetic perturbation reagents (shRNAs or CRIPR/Cas9) to silence or knock-out individual genes from the Achilles project[10].

Multidisciplinary integration for Responsible Research and Innovation (RRI)

The innovative and transformative powers of science can be studied in terms of how actions are mediated through the field’s experimental systems. In Rheinberger’s words, an experimental system is the smallest working unit designed to give unknown answers to questions that the experimenters are not yet able clearly to ask. The process of constructing these systems is governed by an internal dynamics, what Hacking described as the “self-vindicating” dynamics of laboratory research. Scientific work, when succeeding, may be controversial as it involves building machineries for creating common futures, which is particularly evident in large experimental systems built to enable innovation[11].

The notion of experimental systems identifies the task of RRI research initiatives in the context of the need to “rethink science”, or the call for “new social contracts”[12]. A conceptual and institutional ideals of clear separation between scientific and societal activities that have governed professional divisions of labour. These orders have been identified and discussed in terms of the “social contract” between science and society, ideals that have been argued as necessary to be reconsidered in light of what science has become[13].

The smallest working units of science is no longer easily confined to a laboratory or a research group, like exemplified in the work needed to build research infrastructures enabling innovations pathways for precision oncology. The prototype system engineered in the project provide a platform for they study of enabling innovation systems of the future. The hallmark of experimental sciences can be seen as the one of creating orders or “stability”[14]. Such stable or reliable orders is arguably in large research structures more evidently crossing over the natural and the social, often referred to as “socio-technical” orders. Building such orders is now also explicitly expressed as goals of large scale scientific initiatives (typically labelled as enabling or converging technologies). RRI/ELSA initiatives emerged in the context of such priority areas where changes in modern science are particularly evident (widely discussed as shifts from normal to “post-normal”, academic to ”post-academic” or Mode 1 to “Mode2” science). RRI initiatives reflect how scientific activities are seen to be a collective social concern as they perform “collective experiments” on our common futures, that in turn call rethinking ways to coordinate scientific, industrial and societal efforts[15].

RRI activities are widely regarded as urgently needed, despite of few generally recognised success stories and lack of unifying analysis across sectors of the why’s and how’s of RRI. State-of-the-art RRI actions are basically still at a stage of outlines of frameworks and definitions[16]. In our analysis, the work of restructuring normative orders are critically challenging for RRI initiatives. The challenge of RRI thus needs to be understood in terms of how professional identities and goals are challenged. The very understanding of the ethos of one’s professional practice includes how it is to be conducted well in relation to other adjacent practices. Integrated projects appear to us as one, among many, important RRI approaches as they provide an important venue for engaging the ethos of collectives as well as constituent fields of practitioners.

International collaborations

For technology development and implementation we collaborate with:

Denis Thieffry – PSL Université – Ecole Normale Supérieure, IBENS/INSERM, Paris.

Laurence Calzone – Computational Systems Biology of Cancer, Institut Curie, Paris.

Vincent Noël – Research Engineer at Institut Curie, Paris.

Anna Niarakis – University of Evry Val d’Essonne, Paris-Saclay.

Julio Saez-Rodriguez – University of Heidelberg.

[1] Al-Lazikani et al Combinatorial drug therapy for cancer in the post-genomic era. Nat Biotechnol 30: 679–, 2012.
[2] Kauffman S. Homeostasis and differentiation in random genetic control networks. Nature 1969, 224:177–8.
[3] Thomas R. Boolean formalisation of genetic control circuits. J Theor Biol 1973, 42:565–583.
[4] Wolkenhauer O.  Front Physiol. 5:21. 2014. PMID: 24478728; Wolkenhauer O, et al..Genome Med. 26:21-, 2014. PMID: 25031615.
[5] Naldi A, Remy E, Thieffry D, Chaouiya C. Dynamically consistent reduction of logical regulatory graphs. Theor Comput Sci. 2011;412: 2207-, 2011; Grieco L, Calzone L, …, Thieffry D. Integrative modelling of the influence of MAPK network on cancer cell fate decision. Miyano S, editor. PLoS Comput Biol. 9: e1003286-, 2013.
[6] Green, S., & Wolkenhauer, O. (2012). Integration in action. EMBO reports, 13(9), 769-771.
[7] Leitner F, Krallinger M, Tripathi S, Kuiper M, Lægreid A, Valencia A. Mining cis-Regulatory Transcription Networks from Literature. Proceedings of BioLINK, ISMB/ECCB SIG 2013.
[8] Leitner F, et al Nat Biotechnol. 2010 28:897-. PMID: 20829821; Salgado D, et al .Bioinformatics. 2012, 28:2285- PMID: 22789588.
[9] Bansal M, et al. . A community comp. challenge to predict the activity of pairs of compounds. Nat Biotechnol  32:1213-, 2014.
[10] TCGA: http://cancergenome.nih.gov/; ICGC: https://dcc.icgc.org/); CCLE (http://www.broadinstitute.org/ccle; Achilles: http://www.broadinstitute.org/achilles).
[11] Hacking,I. 1992 The Self-Vindication of the Laboratory Sciences. In Pickering 1992 (ed.) Science as Practice and Culture. – Rheinberger, H-J.1997 Toward a History of Epistemic Things – Rabinow,P.et.al.2005 A Machine to make a future.
[12] Nydal,R.2005 Rethinking the topoi of normativity. Phil. dissertation, NTNU – Nowtny, H. et.al.2001 Re-thinking science.
[13] Winner, L. 1993 A New Social Contract for Science. Technology Review (96) 65. – Lubchenco, J. 1997 Entering the Century of the Environment: A New Social Contract for Science. Science (279) 491. – Guston, D. H. and Keniston, K. 1994 Updating the Social Contract for Science. Technology Review (97) 60. – Gibbons, M. 1999 Science’s New Social Contract With Society. Nature (402) 81.
[14] Fujimura, J. H. 1996.  Crafting Science. Galison, P. 1987. How Experiments End.  Pickering, A. 1995  The Mangle of Practice. Rheinberger, H-J. 1997. Toward a History of Epistemic Things. Knorr-Cetina, K. 1999 Epistemic Cultures.
[15] Latour, B 2004 “Which protocol for the new collective expeiments?” In Schmindgen,H. (ed) Experimental cultures  European Commission reports a) Nordmann A. 2004 Converging Technologies – Shaping the Future of Euopean Societies.b)Hoven J 2013 Options for streghtening responsible research and innovation.  Latour, B. 2004 The Politics of Nature.
[16] Rip, A.; Schot, J.W.; Misa, T.J. 1995  Managing Technology in Society. The Approach of Constructive Technology Assessment – Guston, D, og Sarewitz, D. 2001 Real-time technology assessment. Science and Public Policy (33) 5-16, 2001 –  Conferansen I rom, Rip, RRI boken.