Search

Prostate cancer is the second most occurring cancer in men worldwide. To better understand the mechanisms of tumorigenesis and possible treatment responses, we developed a mathematical model of prostate cancer which considers the major signalling pathways known to be deregulated. We personalised this Boolean model to molecular data to reflect the heterogeneity and specific response to perturbations of cancer patients. 488 prostate samples were used to build patient-specific models and compared to available clinical data. Additionally, eight prostate cell-line-specific models were built to validate our approach with dose-response data of several drugs. The effects of single and combined drugs were tested in these models under different growth conditions. We identified 15 actionable points of interventions in one cell-line-specific model whose inactivation hinders tumorigenesis. To validate these results, we tested nine small molecule inhibitors of five of those putative targets and found a dose-dependent effect on four of them, notably those targeting HSP90 and PI3K. These results highlight the predictive power of our personalised Boolean models and illustrate how they can be used for precision oncology. This work has been partially supported by the European Commission under the PrECISE project (H2020-PHC-668858), the INFORE project (H2020-ICT-825070) and the PerMedCoE (H2020-ICT-951773) Peer Reviewed "Article signat per 12 autors/es: Arnau Montagud, Jonas Béal, Luis Tobalina, Pauline Traynard, Vigneshwari Subramanian, Bence Szalai, Róbert Alföldi, László Puskás, Alfonso Valencia, Emmanuel Barillot, Julio Saez-Rodriguez, Laurence Calzone"

Multiplex heterogeneous networks (networks with multiple interconnected layers) could provide significant information to understand the global topology, when compared to the networks that rely on single type of information Random Walk with Restart on Multiplex Heterogeneous networks (RWR-MH) have been shown to effectively identify candidate genes of not well-studied diseases Random walk with restart algorithm allows to explore all neighbours in close proximity to the seed nodes (diseases/ genes/ drugs)

Designing drugs that are selective is crucial in pharmaceutical research to avoid unwanted side effects. To decipher selectivity of drug targets, computational approaches that utilize the sequence and structural information of the protein binding pockets are frequently exploited. In addition to methods that rely only on protein information, quantitative approaches such as proteochemometrics (PCM) use the combination of protein and ligand descriptions to derive quantitative relationships with binding affinity. PCM aims to explain cross-interactions between the different proteins and ligands, hence facilitating our understanding of selectivity. The main goal of this dissertation is to develop and apply field-based PCM to improve the understanding of relevant molecular interactions through visual illustrations. Field-based description that depends on the 3D structural information of proteins enhances visual interpretability of PCM models relative to the frequently used sequence-based descriptors for proteins. In these field-based PCM studies, knowledge-based fields that explain polarity and lipophilicity of the binding pockets and WaterMap-derived fields that elucidate the positions and energetics of water molecules are used together with the various 2D / 3D ligand descriptors to investigate the selectivity profiles of kinases and serine proteases. Field-based PCM is first applied to protein kinases, for which designing selective inhibitors has always been a challenge, owing to their highly similar ATP binding pockets. Our studies show that the method could be successfully applied to pinpoint the regions influencing the binding affinity and selectivity of kinases. As an extension of the initial studies conducted on a set of 50 kinases and 80 inhibitors, field-based PCM was used to build classification models on a large dataset (95 kinases and 1572 inhibitors) to distinguish active from inactive ligands. The prediction of the bioactivities of external test set compounds or kinases with accuracies over 80% (Matthews correlation coefficient, MCC: ~0.50) and area under the ROC curve (AUC) above 0.8 together with the visual inspection of the regions promoting activity demonstrates the ability of field-based PCM to generate both predictive and visually interpretable models. Further, the application of this method to serine proteases provides an overview of the sub-pocket specificities, which is crucial for inhibitor design. Additionally, alignment-independent Zernike descriptors derived from fields were used in PCM models to study the influence of protein superimpositions on field comparisons and subsequent PCM modelling. Lääketutkimuksessa selektiivisten lääkeaineiden suunnittelu on ratkaisevan tärkeää haittavaikutusten välttämiseksi. Kohdeselektiivisyyden selvittämiseen käytetään usein tietokoneavusteisia menetelmiä, jotka hyödyntävät proteiinien sitoutumiskohtien sekvenssi- ja rakennetietoja. Proteiinilähtöisten menetelmien lisäksi kvantitatiiviset menetelmät kuten proteokemometria (proteochemometrics, PCM) yhdistävät sekä proteiinin että ligandin tietoja muodostaessaan kvantitatiivisen suhteen sitoutumisaffiniteettiin. PCM pyrkii selittämään eri proteiinien ja ligandien vuorovaikutuksia ja näin auttaa ymmärtämään selektiivisyyttä. Väitöstutkimuksen tavoitteena oli kehittää ja hyödyntää kenttäpohjaista proteokemometriaa, joka auttaa ymmärtämään relevantteja molekyylitasoisia vuorovaikutuksia visuaalisen esitystavan kautta. Proteiinin kolmiulotteisesta rakenteesta riippuva kenttäpohjainen kuvaus helpottaa PCM-mallien tulkintaa, etenkin usein käytettyihin sekvenssipohjaisiin kuvauksiin verrattuna. Näissä kenttäpohjaisissa PCM-mallinnuksissa käytettiin tietoperustaisia sitoutumistaskun polaarisuutta ja lipofiilisyyttä kuvaavia kenttiä ja WaterMap-ohjelman tuottamia vesimolekyylien sijaintia ja energiaa havainnollistavia kenttiä yhdessä lukuisten ligandia kuvaavien 2D- ja 3D-deskriptorien kanssa. Malleja sovellettiin kinaasien ja seriiniproteaasien selektiivisyysprofiilien tutkimukseen. Tutkimuksen ensimmäisessä osassa kenttäpohjaista PCM-mallinnusta sovellettiin proteiinikinaaseihin, joille selektiivisten inhibiittorien suunnittelu on haastavaa samankaltaisten ATP sitoutumistaskujen takia. Tutkimuksemme osoitti menetelmän soveltuvan kinaasien sitoutumisaffiniteettia ja selektiivisyyttä ohjaavien alueiden osoittamiseen. Jatkona 50 kinaasia ja 80 inhibiittoria käsittäneelle alkuperäiselle tutkimukselle rakensimme kenttäpohjaisia PCM-luokittelumalleja suuremmalle joukolle kinaaseja (95) ja inhibiittoreita (1572) erotellaksemme aktiiviset ja inaktiiviset ligandit toisistaan. Ulkoisen testiyhdiste- tai testikinaasijoukon bioaktiivisuuksien ennustaminen yli 80 % tarkkuudella (Matthews korrelaatiokerroin, MCC noin 0,50) ja ROC-käyrän alle jäävä ala (AUC) yli 0,8 yhdessä aktiivisuutta tukevien alueiden visuaalisen tarkastelun kanssa osoittivat kenttäpohjaisen PCM:n pystyvän tuottamaan sekä ennustavia että visuaalisesti ymmärrettäviä malleja. Tutkimuksen toisessa osassa metodin soveltaminen seriiniproteaaseihin tuotti yleisnäkemyksen sitoutumistaskun eri osien spesifisyyksistä, mikä on ensiarvoisen tärkeää inhibiittorien suunnittelulle. Lisäksi kentistä johdettuja, proteiinien päällekkäinasettelusta riippumattomia Zernike-deskriptoreita hyödynnettiin PCM-malleissa arvioidaksemme proteiinien päällekkäinasettelun vaikutusta kenttien vertailuun ja sen jälkeiseen PCM-mallinnukseen.

In line with recent advances in neural drug design and sensitivity prediction, we propose a novel architecture for interpretable prediction of anticancer compound sensitivity using a multimodal attention-based convolutional encoder. Our model is based on the three key pillars of drug sensitivity: compounds’ structure in the form of a SMILES sequence, gene expression profiles of tumors, and prior knowledge on intracellular interactions from protein–protein interaction networks. We demonstrate that our multiscale convolutional attention-based encoder significantly outperforms a baseline model trained on Morgan fingerprints and a selection of encoders based on SMILES, as well as the previously reported state-of-the-art for multimodal drug sensitivity prediction (R2 = 0.86 and RMSE = 0.89). Moreover, the explainability of our approach is demonstrated by a thorough analysis of the attention weights. We show that the attended genes significantly enrich apoptotic processes and that the drug attention is strongly correlated with a standard chemical structure similarity index. Finally, we report a case study of two receptor tyrosine kinase (RTK) inhibitors acting on a leukemia cell line, showcasing the ability of the model to focus on informative genes and submolecular regions of the two compounds. The demonstrated generalizability and the interpretability of our model testify to its potential for in silico prediction of anticancer compound efficacy on unseen cancer cells, positioning it as a valid solution for the development of personalized therapies as well as for the evaluation of candidate compounds in de novo drug design., Molecular Pharmaceutics, 16 (12), ISSN:1543-8384, ISSN:1543-8392

PaccMann tackles the challenging problem of drug sensitivity prediction adopting a holistic approach. The model was trained on data from Genomics of Drug Sensitivity in Cancer (GDSC, https://www.cancerrxgene.org/)