AI for Life Scientists: from Basics to ApplicationsOnline Event

11 Apr 2024, 14:00 → 25 Jul 2024, 15:30 Europe/Berlin

Zoom

Axel Schumacher (KIT), Donatella Cea (Helmholtz AI), Lisa Barros de Andrade e Sousa (Helmholtz AI), Nicole Merkle (KIT), Sikha Ray (KIT), Thorsten Auth (Forschungszentrum Jülich)

Artificial Intelligence for Life Scientists

From Basics to Applications

In this lecture series, the use of artificial intelligence in life sciences will be discussed guided by applications in omics and drug discovery. The lecturers will introduce you to basic concepts like regression and classification, taking into account typical challenges like data scarcity that is often an issue for biological data, and potential biases and the black-box nature of machine-learning approaches. An introduction to supervised and unsupervised learning will encompass linear and tree-based approaches, ensemble methods, and neural networks for supervised learning, as well as clustering and dimensionality reduction for unsupervised learning. You will learn more about omics data including related experimental techniques and inherent biases, as well as data preprocessing, such as data normalization and dealing with class imbalance. The techniques of generative AI and the principles of diffusion models will be illustrated for small-molecule generation and protein-ligand interaction; transformer models will be discussed specifically for DNA sequence analysis. Model-agnostic and model-specific methods for Explainable AI will be discussed, highlighting the importance of understanding model predictions to reveal biological mechanisms. Finally, artificial-intelligence applications in drug discovery and (gen)omics along with their potential importance for developing new therapies and personalized medicine will be discussed along with ethical decision-making. 

 

 

Dates: Thursdays, 14:00-15:30.

Information on the dates, speakers and lecture titles can be found below.

Organization & Registration

This online lecture series is held by experts from and invited by the organizing schools and Helmholtz AI. It covers various aspects, from basics to applications. Therefore, it is recommended that you attend all lectures for a deep understanding of the subject. Registration is required to attend. Places are limited, and priority will be given to fellows (members) of the three schools in case of overbooking. Participants with an attendance rate of more than 70% may receive a certificate of attendance.

Please register here.

---

This event is organized by

International Helmholtz Research School of Biophysics and Soft Matter    |    BioInterfaces International Graduate School

Helmholtz Information & Data Science School for Health    |    Helmholtz AI

Vortragsankuendigung_240411.pdf

Thursday 11 April

Lecture Series: Overview

Conveners: Axel Schumacher (KIT), Donatella Cea (Helmholtz AI)

1 Introduction to AI in Life Sciences

The talk will provide an introduction to Artificial Intelligence (AI) applications in life sciences. Beginning with fundamental concepts such as regression and classification tasks, it will explain the key metrics used to assess model performance and showcase specific applications in life sciences. The lecture will also address the challenges and limitations of AI models, like data scarcity, potential biases, and AI’s ‘black-box’ nature. Additionally, it explores the integration of AI in life sciences, discussing its potential in advancing research and clinical solutions.

Speaker: Marie Piraud (Helmholtz AI, Helmholtz Munich)

Thursday 18 April

Lecture Series: Supervised and Unsupervised Learning

Conveners: Axel Schumacher (KIT), Sikha Ray (KIT)

2 Supervised and Unsupervised Learning

This talk provides an overview of key concepts in machine learning, focusing on the distinctions and applications of Supervised and Unsupervised learning. It dives into various supervised model types, including linear, tree-based, and ensemble approaches and neural networks, but also touches on unsupervised methods like clustering and dimensionality reduction. The discussion encompasses the fundamentals of these models, their applications, and key differences, providing a comprehensive overview suitable for understanding various machine learning approaches.

Speaker: Axel Loewe (KIT, IBT)

Thursday 25 April

Lecture Series: Introduction to Biological Data

Conveners: Thorsten Auth (Forschungszentrum Jülich), Nicole Merkle (KIT)

3 Biological Data

This talk provides an introduction to omics data, highlighting the different omics data types, such as genomics and proteomics, and explaining experimental methods like the collection of transcriptomics data via RNA sequencing. In addition, it addresses the challenges of collecting omics data and their inherent biases. The aim is to provide a fundamental understanding of omics data characteristics, which is essential for applying machine learning in omics research and applications.

Speaker: Björn Usadel (Forschungszentrum Jülich, IBG)

Thursday 2 May

Lecture Series: Data Preprocessing

Conveners: Thorsten Auth (Forschungszentrum Jülich), Donatella Cea (Helmholtz AI)

4 Data Preprocessing

This talk will give a general overview of data analysis steps commonly preceding the application of actual AI modeling approaches. Possible challenges and solutions in handling real-world data will be illustrated with examples of omics datasets. After highlighting the importance of data representation, normalization, quality control and batch effect correction, the lecture will address how to deal with missing values, outliers, class imbalance and different data types. Furthermore, a main focus will be on dimensionality reduction techniques, including methods for feature transformation, feature aggregation and feature selection.

Speaker: Elisabeth Georgii (Helmholtz AI, Helmholtz Munich)

Thursday 23 May

Lecture Series: CANCELED: Diffusion Models for Biomedical Applications

Conveners: Nicole Merkle (KIT), Thorsten Auth (Forschungszentrum Jülich)

5 Diffusion Models

Unfortunately, the lecture by Bastian Riek planned for 23 May has to be rescheduled. We will inform all participants registered for the lecture series by email about a new date.

Diffusion models are a novel class of generative machine learning models that have recently gained significant traction in different fields, including biomedical applications. Such models operate by gradually injecting additional noise into data, yielding a distribution of random noise. To generate new unseen instances, the random noise distribution is sampled and subjected to a denoising step. This process results in a new instance from the target distribution. In the context of biomedical applications, diffusion models hold promise for a wide range of different tasks, from drug discovery to medical imaging analysis. This talk provides a brief overview of their potential and their challenges for large-scale biomedical applications.

Speaker: Bastian Rieck (Helmholtz Munich, Technical University of Munich)

Thursday 13 June

Lecture Series: Transformer Language Models for Genomic Sequences

Conveners: Sikha Ray (KIT), Nicole Merkle (KIT)

6 Transformer Models

The rapid evolution of transformer-based large language models (LLMs) has revolutionized numerous fields, with their impact on genomics promising to unlock new frontiers in biology and medicine. In this lecture, we will cover basics of language models, explaining how they predict and interpret sequences of symbols, and how we process DNA sequences for computational processing. The core of the lecture will delve into transformer models, emphasizing their self-attention mechanisms and their ability to handle long-range dependencies in data, making them ideal for analyzing the complex sequences found in DNA and RNA. By the end of the lecture, attendees will understand the inner workings of the transformer architecture and the application of such language models in genomics.

Speaker: Vasikmil Martinek (CEITEC MU)

Thursday 27 June

Lecture Series: Ethical Considerations for AI Models in the Life Sciences

Conveners: Thorsten Auth (Forschungszentrum Jülich), Donatella Cea (Helmholtz AI)

7 Ethical Considerations for AI Models

This lecture provides a comprehensive overview of the ethical landscape surrounding AI models. We will briefly cover the history of AI ethics to understand concepts particularly relevant in the life sciences, such as non-maleficence, bias, fairness, transparency, and responsibility. From there, we will look at life-science-relevant AI cases to discuss the potential real-world implications of these concepts. We will round up the session by discussing state-of-the-art tools to mitigate ethical challenges when creating AI in the life sciences.

Speaker: Theresa Willem (Helmholtz Munich, Institute of History and Ethics in Medicine, Department of Clinical Medicine, TUM School of Medicine and Health)

Thursday 4 July

Lecture Series: Explainable AI

Conveners: Nicole Merkle (KIT), Sikha Ray (KIT)

8 Explainable AI

This talk dives into Explainable AI (XAI), exploring its significance in demystifying complex Machine Learning (ML) and Deep Learning (DL) models. It addresses the necessity of understanding model predictions, especially in biology, to reveal mechanisms behind biological systems. The talk covers model-agnostic and model-specific methods, highlighting the challenge of using XAI methods when features are not independent. It emphasises how XAI contributes to scientific understanding, safety, and ethical considerations, ensuring models are accurate but also fair and trustworthy

Speaker: Scott Thiebes (KIT)

Thursday 11 July

Lecture Series: CANCELED: Generative Modeling for Phenotypic Based Drug Discovery

Conveners: Thorsten Auth (Forschungszentrum Jülich), Nicole Merkle (KIT)

9 AI for Drug Discovery

This lecture introduces drug discovery, emphasising the role of Machine Learning (ML) and Deep Learning (DL) in modern pharmaceutical research. It covers the application of these technologies in predicting drug-target interactions, highlighting their importance in developing new therapies. The session aims to offer a comprehensive overview of the intersection between AI and drug discovery, focusing on its practical implications in healthcare.Lecture 10: AI for Genomics

Speaker: Mohammed Lotfollahi (Wellcome Sanger Institute, CCAIM, University of Cambridge)

Monday 15 July

Lecture Series: Diffusion Models for Biomedical Applications

Conveners: Nicole Merkle, Thorsten Auth (Forschungszentrum Jülich)

10 Diffusion Models

Diffusion models are a novel class of generative machine learning models that have recently gained significant traction in different fields, including biomedical applications. Such models operate by gradually injecting additional noise into data, yielding a distribution of random noise. To generate new unseen instances, the random noise distribution is sampled and subjected to a denoising step. This process results in a new instance from the target distribution. In the context of biomedical applications, diffusion models hold promise for a wide range of different tasks, from drug discovery to medical imaging analysis. This talk provides a brief overview of their potential and their challenges for large-scale biomedical applications.

Speaker: Bastian Rieck (Helmholtz Munich, Technical University of Munich)

Thursday 18 July

Lecture Series: AI for Omics

Conveners: Axel Schumacher (KIT), Sikha Ray (KIT)

11 AI for Omics

This lecture introduces applications of AI models to omics data in the context of precision medicine. The key goals are 1) to understand the molecular processes underlying complex human phenotypes to be able to rationally device new therapeutic strategies; 2) to identify individuals at high risk of disease and 3) to predict the response to treatment. The lecture will introduce recent work from the Heinig lab addressing these aims for cardiovascular diseases.

Speaker: Matthias Heinig (Helmholtz Munich, TUM School of Computation, Information and Technology)

Thursday 25 July

Lecture Series: ChatGPT in Action: Enhancing your Workflow

Conveners: Sikha Ray (KIT), Thorsten Auth (Forschungszentrum Jülich)

12 ChatGPT in Action

This lecture provides a high-level overview of the capabilities and future perspectives of ChatGPT and GPT-4, focusing on their applications in the scientific world. It provides insights into the workings of large language models (LLMs) and offers practical guidance on how to use ChatGPT as a scientist effectively. Attendees will learn about enhancing their research and writing through advanced prompt engineering techniques. The session also covers best practices in coding with ChatGPT with practical examples.

Speakers: Donatella Cea (Helmholtz AI, Helmholtz Munich), Ilhem Isra Mekki (Helmholtz AI, Helmholtz Munich)