To view previous webinars use the links below
2020 Webinars | 2021 Webinars | 2022 Webinars | 2023 Webinars
ISCBacademy is an online webinar series including the ISCB COSI, COVID webinars, Indigenous Voices and practical tutorials. We aim to inspire, connect, and communicate the science while providing a hands-on experience accessing and using newly developed bioinformatics tools while ensuring best practices for rigour and reproducibility.
May 30, 2023 at 4:00 PM CET
Complex supervised Machine Learning (ML) and Deep Learning (DL) models are often considered to be “Black Boxes” because it can be hard to understand why certain predictions have been made by the model. Particularly in biology, it is more and more important to not just accurately predict the outcome of a biological system with a machine learning model but to also be able to uncover the mechanisms behind those biological systems that led to a certain outcome. To uncover the underlying mechanisms of a biological system, we have to work on the interpretability of our models, which are able to learn the underlying patterns in our data. Besides this important scientific aspect, interpretability can play an important role in addressing problems related to safety and ethics. To deploy our models to the real world, it is indispensable to not only make accurate predictions, but also to understand the logic behind those predictions. Only by understanding the model’s decision-making process, we can ensure that the model produces valuable insights. Indeed, the application of interpretability methods allows us to identify potential biases, ensure fairness and prevent genre, social, and race discrimination. Moreover, in many fields, but in particular in the biomedical domain, explainability can be the key to acceptance and trust.
During this 2-hour tutorial on eXplainable Artificial Intelligence (XAI), we want to identify the main aspects of why interpretability – in an era where AI is becoming more and more pervasive – is so important. After this general introduction, we will focus on model-agnostic state-of-the-art methods like Permutation Feature Importance and SHAP, as well as model-specific methods like Forest-Guided Clustering. During the practical sessions, the participants can discuss those methods with peers and mentors, and get familiar with the most popular python libraries on XAI.
This event has been cancelled.
Hosted by:
June 2, 2023 at 11:00 AM EDT
Non–small cell lung cancer (NSCLC) is the leading cause of cancer related death worldwide and a complex disease where multiple treatment options are required to address the needs of different patient populations. Significant progress has been made in the first-line treatment for patients with advanced NSCLC, ineligible for targeted therapy, by the investigation of checkpoint pathway blockade. Such immunotherapy (I-O)-based treatment regimens have been assessed as a monotherapy in patients whose tumors express programmed death ligand 1 (PD-L1), as well in combination with chemotherapy, regardless of PD-L1 expression.1-12 While durable responses and prolonged survival have been demonstrated in some patients treated with I-O, there remains a high disease burden and a need to improve the ability to predict which patients are more likely to receive benefit from treatment with I-O.
CheckMate 026 is a phase III, international, randomized, open-label trial comparing the efficacy and safety of single-agent nivolumab with those of platinum-based chemotherapy as first line therapy in patients with NSCLC and tumor PD-L1 ≥1%. While nivolumab had a favorable safety profile over chemotherapy, progression free survival (PFS) was not significantly longer with nivolumab and overall survival was similar between treatment groups. Owing to the complexity of the immune system, identification of novel biomarkers of response to I-O–based treatments are still required in order to better predict survival and durable responses for subset of NSCLC patients. While immunostaining for PD-L1 expression is currently used to predict those patients likely to respond to I-O, several challenges remain with accurate detection and scoring of PD-L1. This Anti-PD-1 Response DREAM Challenge is a crowdsourced effort aiming to identify predictive biomarkers of response to anti-PD-1 monotherapy (nivolumab), in patients with NSCLC using clinical data and gene expression data from the phase 3 trial, CheckMate 026. This trial is an international, randomized, open-label trial comparing the efficacy and safety of single-agent nivolumab with those of platinum-based chemotherapy as first line therapy in patients with NSCLC. Overall survival was similar and progression-free survival was not improved, however, nivolumab had a favorable safety profile compared with chemotherapy. Owing to the complexity of the immune system, novel biomarkers for response are being explored and may improve survival for a subset of NSCLC patients, while also facilitating better patient stratification to assist in the development of novel therapeutic approaches for NSCLC patients. While immunostaining for PD-L1 expression is currently used to identify patients eligible for I-O monotherapy, several challenges remain with accurate detection and scoring of PD-L1. This Anti-PD-1 Response DREAM Challenge is a crowdsourced effort aiming to identify predictive biomarkers of response to anti-PD-1 monotherapy (nivolumab), in patients with metastatic NSCLC using clinical data and gene expression data from CheckMate 026.
Join the Webinar by logging in to ISCB Nucleus
Hosted by:
June 20, 2023 at 4:00 PM BST
ISCB’s 3DSIG, in collaboration with Elixir 3D-Biodata Community, is pleased to announce this upcoming webinar. This webinar will have two speakers. Detailed information on the program and how to join the session are located at https://elixir-europe.org/events/3d-bioinfo-webinar-protein-design-and-evolution
Hosted by:
June 27, 2023 at 11:00 AM EDT
Recent advances in machine learning have opened new paths for making predictions based on large sets of complex life science data. However, is it also possible to gain an understanding of the complex underlying systems? And, is this possible when there is only a small number of labelled samples available?
Firstly, we demonstrate how to analyse the impact of a genomic alteration using system level response data. We employed machine learning methodology to explore the problem and subsequently developed a statistical approach to analyse the impact of genomic alterations using as few as 20 samples. Applying this approach, we identified several novel structural variants that are likely to have a significant impact on the development of colorectal cancer.
In a second example, we demonstrate that it is possible to predict which proteins are likely to be found in extracellular vesicles (EVs). By using meaningful input features, we can gain an understanding of both the vesicle-sorting mechanism, as well as understanding why specific proteins are likely to be found in vesicles. In addition, we can shed light on which experimental EV extraction protocols are most reliable. Finally, I illustrate how we can tackle similar questions - that have a limited number of labelled data - by using multi-task deep learning architectures that can be enhanced by using labelled data from related problems.
Join the Webinar by logging in to ISCB Nucleus
Hosted by: