Real-world data on the clinical utility of a novel artificial intelligence-based computational method to support treatment decisions in gastrointestinal cancers

e13616

Background: Several molecularly targeted agents (MTA) are registered in the same indication, even in the same line, to treat gastrointestinal (GI) cancers. Comprehensive molecular profiling is readily available for many patients, but only a few biomarkers are used to make clinical decisions between on-label MTAs. Most tumors harbor a combination of driver alterations potentially affecting the sensitivity to MTAs according to hundreds or thousands of evidence items that can explain the heterogeneous clinical response of individual patients to these therapies. We have developed an AI-based computational method, an automated, evidence-based reasoning framework, called digital drug assignment (DDA) to provide decision support based on the complete molecular profile for each patient. DDA has been shown to potentially improve treatment decisions in case of complex molecular profiles acquired in the SHIVA01 trial. In the present study, we analyzed the clinical utility of DDA in on-label treatment decisions in GI cancers. Methods: We analyzed the next-generation sequencing (NGS) data obtained using 50-591-gene panels of 239 patients with GI tract cancers from our everyday practice. Molecular profiles were uploaded to a DDA-based software system to calculate the aggregated evidence level (AEL) values of associated MTAs registered in the same indication of selected GI cancers. Decision support rate (DSR) was calculated as the percentage of patients where DDA could support clinical decisions based on the presence of genetic alterations associated with any of the available MTAs. Results: Targeted treatment options registered in the same indications and the DDA-based decision support rates (DSR) are shown in the table. For example, DDA could support the decision between cetuximab and bevacizumab based on NGS results in 47% of RAS/RAF wild-type colorectal cancer cases. Larger NGS panels yielded higher rates of decision support (44% vs 52% with gene panels under and over 300 genes, respectively). Conclusions: DDA-based computational reasoning can support molecular profile-based treatment decisions among different on-label MTAs in a clinically meaningful fraction of GI cancers.[Table: see text]

Performance analysis of a novel artificial intelligence-based computational method on published ex vivo drug sensitivity data to support targeted treatment decisions in precision oncology

e13618

Background: Comprehensive molecular profiling is readily available for clinical practice. An extensive amount of published evidence provides information about the potential functional relevance of many driver genes and genetic alterations. But, due to the large number of driver genes and alterations, and the long-tail frequency distribution of these alterations and their possible combinations, very few single driver alterations have reached high-enough-level evidence alone to support clinical decisions. We hypothesized that by aggregating evidence following logical principles of reasoning by a computational system, this vast molecular information can be used to improve personalized treatment decisions. In the present study, we used previously published ex vivo drug sensitivity data to analyze the performance of an artificial intelligence-based computational method, the digital drug assignment (DDA), which has shown utility by improving treatment decisions in case of complex molecular profiles acquired in the SHIVA01 trial. Methods: We selected 111 cases with whole-genome sequencing (WGS) and ex vivo drug sensitivity data of a previously published acute myeloid leukemia study (Tyner et al, 2018). WGS variants were filtered for a preselected hematology-related panel of 446 genes and uploaded to a DDA-based software system to calculate the aggregated evidence level (AEL) values of associated molecularly targeted agents. DDA-predicted sensitivity (or resistance) was defined as AEL > 0 (or AEL < 0) in the presence of at least one actionable driver. Area under the curve (AUC) values were used for determining the ex vivo sensitivity or resistance of leukemia cells to 40 approved drugs and 53 developmental compounds. Results: The AUC values were significantly different in the drug-sensitive and -resistant groups forecasted by the DDA (167.1 and 205.5, respectively, p < 0.0001) and differed significantly from the average AUC value (194.6). Overall, sensitivity was correctly predicted in 66% of compound-sample pairs (n = 671). 88 approved drugs had AEL value over 1000; of these 73% were effective according to the ex vivo results. While forecasted resistance was confirmed in 63% of the cases. With only the most sensitive/resistant 20% of cases considered from the ex vivo data, sensitivity was accurately predicted in 75% of approved compound-sample pairs (n = 173). 37 approved drugs had AEL value over 1000; of these 81% were confirmed as sensitive. Conclusions: The DDA-based computational reasoning has a promising performance in forecasting sensitivity and resistance to a broad spectrum of targeted agents based on molecular information. Therefore, it has the potential to automate, standardize and improve complex molecular profile-based targeted treatment decisions.

A computational method for prioritizing targeted therapies in precision oncology: performance analysis in the SHIVA01 trial

Precision oncology is currently based on pairing molecularly targeted agents (MTA) to predefined single driver genes or biomarkers. Each tumor harbors a combination of a large number of potential genetic alterations of multiple driver genes in a complex system that limits the potential of this approach. We have developed an artificial intelligence (AI)-assisted computational method, the digital drug-assignment (DDA) system, to prioritize potential MTAs for each cancer patient based on the complex individual molecular profile of their tumor. We analyzed the clinical benefit of the DDA system on the molecular and clinical outcome data of patients treated in the SHIVA01 precision oncology clinical trial with MTAs matched to individual genetic alterations or biomarkers of their tumor. We found that the DDA score assigned to MTAs was significantly higher in patients experiencing disease control than in patients with progressive disease (1523 versus 580, P = 0.037). The median PFS was also significantly longer in patients receiving MTAs with high (1000+ <) than with low (<0) DDA scores (3.95 versus 1.95 months, P = 0.044). Our results indicate that AI-based systems, like DDA, are promising new tools for oncologists to improve the clinical benefit of precision oncology.

Nintedanib Is Active in Malignant Pleural Mesothelioma Cell Models and Inhibits Angiogenesis and Tumor Growth In Vivo

Purpose: Malignant pleural mesothelioma (MPM) is an aggressive thoracic tumor type with limited treatment options and poor prognosis. The angiokinase inhibitor nintedanib has shown promising activity in the LUME-Meso phase II MPM trial and thus is currently being evaluated in the confirmatory LUME-Meso phase III trial. However, the anti-MPM potential of nintedanib has not been studied in the preclinical setting.Experimental Design: We have examined the antineoplastic activity of nintedanib in various in vitro and in vivo models of human MPM.Results: Nintedanib's target receptors were (co)expressed in all the 20 investigated human MPM cell lines. Nintedanib inhibited MPM cell growth in both short- and long-term viability assays. Reduced MPM cell proliferation and migration and the inhibition of Erk1/2 phosphorylation were also observed upon nintedanib treatment in vitro Additive effects on cell viability were detected when nintedanib was combined with cisplatin, a drug routinely used for systemic MPM therapy. In an orthotopic mouse model of human MPM, survival of animals receiving nintedanib per os showed a favorable trend, but no significant benefit. Nintedanib significantly reduced tumor burden and vascularization and prolonged the survival of mice when it was administered intraperitoneally. Importantly, unlike bevacizumab, nintedanib demonstrated significant in vivo antivascular and antitumor potential independently of baseline VEGF-A levels.Conclusions: Nintedanib exerts significant antitumor activity in MPM both in vitro and in vivo These data provide preclinical support for the concept of LUME-Meso trials evaluating nintedanib in patients with unresectable MPM. Clin Cancer Res; 24(15); 3729-40. ©2018 AACR.

The role of Allee effect in modelling post resection recurrence of glioblastoma

Resection of the bulk of a tumour often cannot eliminate all cancer cells, due to their infiltration into the surrounding healthy tissue. This may lead to recurrence of the tumour at a later time. We use a reaction-diffusion equation based model of tumour growth to investigate how the invasion front is delayed by resection, and how this depends on the density and behaviour of the remaining cancer cells. We show that the delay time is highly sensitive to qualitative details of the proliferation dynamics of the cancer cell population. The typically assumed logistic type proliferation leads to unrealistic results, predicting immediate recurrence. We find that in glioblastoma cell cultures the cell proliferation rate is an increasing function of the density at small cell densities. Our analysis suggests that cooperative behaviour of cancer cells, analogous to the Allee effect in ecology, can play a critical role in determining the time until tumour recurrence.

Mathematical models of propagating fronts have been used to represent a wide variety of biological phenomena from action potentials in neural cells to invasive species in ecology and epidemic spreading. Here we show that when such models are used to predict the effects of external perturbations the results can be very sensitive to certain details of the local dynamics. For example, the post resection recurrence of tumour growth depends strongly on the density dependence of the proliferation of cancer cells. This suggests that targeting the cooperative behaviour of cancer cells could be an efficient strategy for delaying the recurrence of diffuse aggressive brain tumours.

Autocrine FGF feedback can establish distinct states of Nanog expression in pluripotent stem cells: a computational analysis

Background

The maintenance of stem cell pluripotency is controlled by a core cluster of transcription factors, NANOG, OCT4 and SOX2 – genes that jointly regulate each other’s expression. The expression of some of these genes, especially of Nanog, is heterogeneous in a population of undifferentiated stem cells in culture. Transient changes in expression levels, as well as heterogeneity of the population is not restricted to this core regulator, but involve a large number of other genes that include growth factors, transcription factors or signal transduction proteins.

Results

As the molecular mechanisms behind NANOG expression heterogeneity is not yet understood, we explore by computational modeling the core transcriptional regulatory circuit and its input from autocrine FGF signals that act through the MAP kinase cascade. We argue that instead of negative feedbacks within the core NANOG-OCT4-SOX2 transcriptional regulatory circuit, autocrine signaling loops such as the Esrrb - FGF - ERK feedback considered here are likely to generate distinct sub-states within the “ON” state of the core Nanog switch. Thus, the experimentally observed fluctuations in Nanog transcription levels are best explained as noise-induced transitions between negative feedback-generated sub-states. We also demonstrate that ERK phosphorilation is altered and being anti-correlated with fluctuating Nanog expression – in accord with model simulations. Our modeling approach assigns an empirically testable function to the transcriptional regulators Klf4 and Esrrb, and predict differential regulation of FGF family members.

Conclusions

We argue that slow fluctuations in Nanog expression likely reflect individual cell-specific changes in parameters of an autocrine feedback loop, such as changes in ligand capture efficiency, receptor numbers or the presence of crosstalks within the MAPK signal transduction pathway. We proposed a model that operates with binding affinities of multiple transcriptional regulators of pluripotency, and the activity of an autocrine signaling pathway. The resulting model produces varied expression levels of several components of pluripotency regulation, largely consistent with empirical observations reported previously and in this present work.

Electronic supplementary material

The online version of this article (doi:10.1186/s12918-014-0112-4) contains supplementary material, which is available to authorized users.