Algorithmic Therapeutic Oncology by Dr. Murat Çobanoğlu, December 24

Topic: Algorithmic Therapeutic Oncology

Speaker: Dr.Murat Çobanoğlu ( UT Southwestern, Lyda Hill Department of Bioinformatics, Texas, USA )

Organized by: Health Informatics Department

Date: 24 December 2018, Monday at 11:30

Place: Graduate School of Informatics Class-01

Abstract: The cost of discovering a new drug has doubled every 9 years to reach more than $2bn per new molecular entity today. The high cost reflects the high rate of failure: only about 10% of all drug candidates in clinical trials get approved. Oncology has the worst success rate among all areas, standing at only 5.1%. Consequently, most new cancer drugs cost in excess of $100k per year per patient, making access to therapy difficult. We want to change that. We think that the solution lies in using clinically relevant assays to replace the reductionist high throughput preclinical assays. The lower throughput of more advanced assays necessitates three or four orders of magnitude reduction in the number of experiments. We propose that large scale public data and machine learning are necessary and sufficient to deliver the required efficiency gain. I will discuss our preliminary results in (i) single cell RNA-seq classification, and clustering using Bayesian Pólya discriminant analysis and Pólya mixture models, (ii) differential TF activity inference using graph constrained fused lasso regularized linear models, (iii) cancer lineage-specific vulnerability identification and exploitation. I will also present one early success story in the form of a drug candidate that specifically kills BRAF WT melanoma, and not normal skin cells or normal immune cells. Drug name and chemical structure will be confidential.

Bio: Murat Can Cobanoglu studied Computer Science & Engineering (major) and Mathematics (minor) in Sabancı University from 2004 to 2008. He then obtained a master’s degree in Computer Science and Engineering also in Sabancı University, between 2008 and 2010, where he worked on the classification of GPCRs using family specific motifs. He received his Ph.D. degree from the Carnegie Mellon University – University of Pittsburgh Joint Ph.D. Program in Computational Biology under the supervision of Ivet Bahar, Ph.D. and D. Lans Taylor, Ph.D. where he worked on computational drug-target interaction prediction and applications in alpha1-antitrypsin deficiency and Huntington’s disease. He then joined the UT Southwestern Medical Center Lyda Hill Department of Bioinformatics as the inaugural UT Southwestern Distinguished Fellow. Outside of work, he is proud to be an épéeist as well as a member of and a frequent visitor to the Dallas Museum of Art.

Announcement Category

Automated Analysis and Synthesis of Facial Actions by Dr. Itır Önal Ertuğrul

Dr. Itır Önal Ertuğrul (Carnegie Mellon University) is going to give a seminar titled "Automated Analysis and Synthesis of Facial Actions". You are all kindly invited.

Location: Conference Room (124), METU Informatics Institute

Time: 28 June 2018, 11:30-12:30

Abstract: Automated analysis of facial actions is a challenging task due to moderate to large head rotation in real-world conditions and differences in the factors of domains such as view angle and resolution. In addition to their automated analysis, synthesis of facial actions has drawn much attention, as it is useful in face data augmentation for emotion recognition. In this talk, I will present 1) a deep approach for multi-label facial action unit detection and its application to assess treatment response in patients treated with deep brain stimulation (DBS) for obsessive-compulsive disorder (OCD); 2) a capsule-based approach to detect multiple action units from facial images having wide range of head orientations; 3) cross-domain generalizability problems of current facial action unit detection models and possible adversarial domain adaptation solutions; and 4) a kinship synthesis framework, in which facial expression videos of probable children are synthesized from videos of parents.

Biography: Itır Önal Ertuğrul received her B.Sc., M.Sc. and Ph.D. degrees from Computer Engineering Department at Middle East Technical University in 2011, 2013 and 2017, respectively. She worked as a Research Assistant at Computer Engineering Department, Middle East Technical University between 2011 - 2017. During her Ph.D., she visited Pattern Recognition and Bioinformatics Group at Delft University of Technology as a visiting Ph.D. student between July – September 2016. After her Ph.D., she worked as a Postdoctoral Researcher at Affect Analysis Group, University of Pittsburgh. Currently, she is a Postdoctoral Researcher at Robotics Institute, Carnegie Mellon University. Her research interests include affective computing, facial expression analysis and biomedical signal processing. For more information, please see her webpage at

Announcement Category

Quantum Computers by Prof. Sadi Turgut

METU Informatics Institute is organizing monthly Friday Seminars series and Prof. Sadi Turgut from METU, Department of Physics is going to give the last talk of this semester. The title of his talk is “Quantum Computers”.

Location: Conference Room (124), METU Informatics Institute

Time: 8 June 2018, 15:30-16:30

This seminar is open to public and those outside METU have to register using Google Forms: (Free Registration)

Short Biography:

Prof. Dr. Sadi Turgut is interested in the physics of information processing, especially quantum information.

BS, 1989, METU Physics,

PhD, 1995, UC Berkeley, Physics,

Working at METU, Physics Department since 1996.

Announcement Category

Personalized Medicine in Cancer Through Bioinformatics- Prof. Dr. Rengül Çetin Atalay

The second talk of METU  Informatics Institute monthly Friday Seminarsseries will be given by Prof. Dr. Rengül Çetin-Atalay from METUInformatics Institute Health Informatics Program:

“Personalized Medicine in Cancer Through Bioinformatics”.

Location: METU Informatics Institute Conference Room (124)

Time: 4 May 2018, 15:30-16:30.

This seminar is open to public and those outside METU have to register from here.

Abstract: Due to the recent advances in molecular biology and the applications of optoelectronic sensors, the volume of biological data in biomedical research has increased significantly. We now acquire large data sets which provides the novel insights for the biology of a disease, in particular for cancer which is recently reported as the most-deadly disease by WHO. The analysis of the molecular characteristics of a patient’s disease through computational tools allows patient stratification for better healthcare which is tailored for personalized prevention, diagnosis, and treatment. In this talk I will summarize recent advances on personalized cancer medicine with a special focus on related KanSiLab studies.

Short Bio: Dr. Rengul Cetin-Atalay holds MD degree and PhD degree in Systems Biology and her research falls in the broad fields of pathway analysis, bioinformatics and biomedical-image analysis and specifically in molecular cellular biology of primary liver cancer. She worked as a research assistant at Ecole Polytechnique, Paris, France during her PhD, and as an assistant professor at Virginia Bioinformatics Institute, Virginia Polytechnic Institute and State University (Virginia Tech), USA during her sabbatical leave in 2004.  She is the recipient of several awards including Turkish Academy of Sciences Young Investigator Award (2002), UICC International Cancer Technology Transfer Award (2008), and Sanovel Pharmaceuticals Inc, Drug Research Award (2013). During the last decade, she extended her studies on the discovery of novel chemotherapeutics for liver cancer by the application of computational techniques on the analysis of pharmaco-transcriptomics data.  Her experience in the interdisciplinary research approach in cancer biology with various aspects of this disease, devotes  her research towards personalized medicine.

Announcement Category

Deep Learning in Computer Vision by Assoc. Prof. Alptekin Temizel

METU Informatics Institute is organizing monthly Friday Seminars series and Assoc. Prof. Alptekin Temizel from METU II Multimedia Informatics Program is going to give the first talk titled “Deep Learning in Computer Vision”.

Deep Learning in Computer Vision

Deep Neural Networks (DNNs) have been shown to achieve human-level accuracy for many vision tasks. On the other hand, they require considerable amounts of computational power and data to train. In this talk, I will first talk about the deep learning based approaches and their differences to traditional approaches. I will then give an overview of the deep learning ecosystem (tools and systems) for training and inference. Then I will be focusing on their application to the computer vision problems with a number of examples including image classification, object detection, image segmentation as well as generative models. I will then conclude by discussing the challenges and problems in current deep learning based approaches.


Alptekin Temizel Assoc. Prof., Graduate School of Informatics, Middle East Technical University (METU) (2007-Current); B.Sc., Electrical and Electronics Engineering, METU (1999); Ph.D. Centre of Vision, Speech and Signal Processing, University of Surrey (2006). He co-founded Visioprime Ltd, UK, a company developing Intelligent Video Systems for Security, where he worked as senior research engineer in 2001-2006 and acquired a number of patents. He started as an Assistant Professor at METU in 2007 where he is currently an Associate Professor. He is the principal investigator of “GPU Education and Research Centre” and NVIDIA Deep Learning Institute Certified Instructor and University Ambassador. He collaborated with and advised several companies on video analytics, video surveillance systems/algorithms and GPU computing. He was a visiting researcher at Microsoft MLDC-Lisbon in the summers of 2014 and 2015. He was on sabbatical leave at University of Birmingham, UK in 2016-2017. He has been expert evaluator of EU H2020 and EUREKA research programs and nationally funded programs. His main research interests are video surveillance, computer vision, machine learning, deep learning, GPU programming and CUDA.

Location: Conference Room (124), METU Informatics Institute
Time: 13 April 2018, 15:30-16:30

This seminar is open to public and those outside METU have to register from EventBrite page of the event: (Free Registration)

Announcement Category

Human Behaviour Analysis in Interactions with Virtual Agents and Robots by Dr. Oya Çeliktutan

Dr. Oya Çeliktutan (Imperial College London) is going to give a seminar titled "Human Behaviour Analysis in Interactions with Virtual Agents and Robots" for Multimedia Informatics - Multimedia Standards (MMI 726) course. You are all kindly invited.

Location: Conference Room (124), METU Informatics Institute
Time: January 5, 2018, 13:30-14:30

Title: Human Behaviour Analysis in Interactions with Virtual Agents and Robots


Socially intelligent agents and robots are soon to be a fact of everyday life due to the rapid growth in the fields of affective computing, human-machine interaction and social robotics. Numerous innovative applications in healthcare, education and entertainment motivate the development of ever more sophisticated machines that are capable of interpreting human behaviours and adapting to user’s profile, namely, personality, preferences and needs. To realise this vision, one of the prominent research problems is automatic personality analysis. Personality is key in communication: people observe others and form impressions regarding their personalities, and adapt their own behaviours to maintain others’ engagement during social interactions. Doing so is a natural routine for humans. In this talk, I will show how this mechanism can be automated for human-virtual agent and human-robot interactions, and I will highlight our research findings on personality and its relationship with engagement. I will continue my talk with a brief introduction to our current work towards deploying such technologies in healthcare, and I will conclude by discussing open problems in the area.


Oya Celiktutan is a postdoctoral researcher in the Personal Robotics Lab in the Department of Electrical and Electronic Engineering at Imperial College London, United Kingdom. Before this, she was a postdoctoral researcher in the Computer Lab at University of Cambridge between 2016 and 2017, and in the School of Electronic Engineering and Computer Science at Queen Mary University of London between 2013 and 2015. She obtained her PhD degree in Electrical and Electronic Engineering from Bogazici University, Turkey, in collaboration with National Institute of Applied Sciences of Lyon, France, in 2013. Her research focuses on computer vision and machine learning within a context of applied work in the areas of human behaviour understanding, affective computing, human-computer interaction and human-robot interaction. She is particularly interested in building smart algorithms to equip machines with the capability of sensing, recognising and interpreting human behaviours for enabling personalised, engaging and assistive interactions with machines.

Announcement Category

Visual Privacy Protection Using False Colors by Asst. Prof. Serdar Çiftçi

Asst. Prof. Serdar Çiftçi (Harran University) is going to give a seminar on Visual Privacy Protection Using False Colors, for Multimedia Informatics - Research Methods (MMI 700) course. You are all kindly invited. Sorry for multiple copies.

Location: Conference Room (124), METU Informatics Institute
Time: January 3, 2018, 11:30-12:30

Visual Privacy Protection Using False Colors


Privacy protection in video surveillance is an important problem, and it will become even more important, as video surveillance is gains in popularity. However, simple methods for protecting privacy are not sufficient as they do not contain all the desired attributes that is expected from a good privacy protection algorithm. Privacy protected images must be reversible if the need arises to view them as unprotected (e.g., during a criminal investigation). Furthermore, protected image should not prohibit non-private statistics to be extracted. Also, the protected content should not be visually disturbing as it is the case with some of the existing privacy protection methods. Perhaps most importantly, the protection must be continuous: that is faces even in a single frame of a video should not remain unprotected. The algorithms that rely on computer vision techniques may therefore be vulnerable to this problem: if an algorithm fails to detect a sensitive region, it will remain unprotected.

In this seminar, methods that are commonly used for visual privacy protection will be reviewed and then a balanced method that was proposed in Ciftci's Ph.D. thesis will be mentioned. The proposed method is reversible, does not create disturbing outputs, do not prohibit collecting non-private information, does not rely on computer vision techniques.


Serdar Ciftci is an Assistant Professor at the Department of Computer Engineering at Harran University. He has obtained his B.Sc. degree from the Department of Computer Engineering at Selcuk University in 2007. He received both his M.Sc. and Ph.D. degrees from the Department of Computer Engineering at Middle East Technical University (METU) respectively in 2001 and 2017. He has been in Multimedia Signal Processing Group Laboratory at École Polytechnique Fédérale de Lausanne (EPFL) as a visiting researcher between January-March 2016. His primary research interest is visual privacy protection. He is a reviewer for various international journals and conferences such as TUBITAK Turkish Journal of Electrical Engineering & Computer Sciences, Selcuk University Journal of Engineering Science and Technology (SUJEST), and ACM Multimedia 2017.

Announcement Category

Social Signal Processing Approaches by Dr. Çiğdem Beyan

Next Wednesday, Dr. Çiğdem Beyan (Istituto Italiano di Tecnologia) is going to give a seminar on Social Signal Processing (SSP) approaches, particularly for emergent leadership, for Multimedia Informatics - Research Methods (MMI 700) course. You are all kindly invited. Sorry for multiple copies.

Location: Conference Room (124), METU Informatics Institute
Time: December 27, 2017, 11:30-12:30


Social interactions are the main facet of human life and also the fundamental research area for social psychology. Even though psychologists have been working on social interactions for a very long time, the automatic analysis of them is a relatively new problem. Social signal processing (SSP) is the field which aims to analyze human interactions in an automatic way using the recent advances in machine analysis (e.g. speech processing, computer vision, machine learning). In this talk, I will survey some approaches and results regarding identification of the social interactions in small group meeting environments particularly for emergent leadership. In detail, a new corpus that has been collected using portable audio and video sensors, novel nonverbal features proposed and the computational methods used will be addressed.


Cigdem Beyan is a postdoctoral researcher at Istituto Italiano di Tecnologia (IIT), Pattern Analysis and Computer Vision (PAVIS) department since September 2015. In PAVIS-IIT, she has been working on social signal processing. She obtained her Ph.D. degree (2015) in School of Informatics, University of Edinburgh, UK, under supervision of Prof. Robert B. Fisher. Her thesis was a part of EU project (FP7) called Fish4Knowledge, in which she mainly investigated fish behaviour understanding. Prior to that, she received her MSc. degree (2010) in School of Informatics, Middle East Technical University where she worked on multiple object tracking and abandoned-object detection utilizing multi-modalities. Beyond her research background, she has a teaching experience (in Turkey and in UK) for 6 years. Additionally, she has been an Associate Fellow of the Higher Education Academy in recognition of attainment against the UK Professional Standards Framework for teaching and learning support in higher education since 2014. She contributed to the computer vision, machine learning and multimedia domains by being (co-) author of more than 25 research papers and has a PCT Patent Application to WIPO. She is a reviewer for various international journals such as Pattern Recognition, IEEE Transactions on Knowledge and Data Engineering and IEEE Transactions on Multimedia and for the conferences such as CVPR, ICCV, BMVC, and ICPR.

Announcement Category

Music Information Retrieval by Dr. Sertan Şentürk

This Wednesday, Dr. Sertan Şentürk (soon to join the R&D team in Kobalt music) is going to give a seminar titled “What is Music Information Retrieval? Applications and the State-of-the-Art” for Multimedia Informatics - Research Methods (MMI 700) course. You are all kindly invited.

Location: Conference Room (124), METU Informatics Institute
Time: December 20, 2017, 11:30-12:30

Abstract: What is Music Information Retrieval? Applications and the State-of-the-Art

Music, without a doubt, is one of the pinnacles of human creativity, communication and culture. With the digital media revolution and the rise of music streaming services, there has recently been a growing interest and a high demand to process large-scale musical material such as audio recordings, music scores and album reviews. In this talk, I will provide a gentle introduction to music information retrieval (MIR), an interdisciplinary field which “incorporates elements from signal processing, machine learning, psychology, musicology, and many more” to retrieve relevant information from music. From recommendation systems to automatic music generation, I will present a wide variety of examples involving MIR and discuss how such technologies are being used in academic and industrial settings, for example in music streaming, computational musicology and music education.

I am a data scientist, specializing in music information retrieval, soon to join the R&D team in Kobalt music. My research involves developing signal processing and machine learning based computational methodologies to automatically describe and discover music. Throughout the years, I have worked on creating robust and scalable solutions for a vast number of tasks such as audio-score alignment, automatic genre/mode classification, music structure analysis, and audio fingerprinting.

In February 2017, I obtained my Ph.D. degree in Information and Communication Technologies from the Music Technology Group in Universitat Pompeu Fabra, Spain. I received my MSc. degree in Music Technology from Georgia Institute of Technology, the USA in 2011, and my BSc. degree in Electrical and Electronics Engineering from Middle East Technical University, Turkey in 2009. I have also completed a four-year, part-time classical guitar program from Hacettepe University Ankara State Conservatory in 2009.

Announcement Category


Subscribe to RSS - Seminar/Talk