• Artificial Intelligence in Medicine

Agenda

This agenda is not final and is subject to change.

  • Day 1Monday, February 18th
  • Day 2Tuesday, February 19th
  • Day 3Wednesday, February 20th
8:00 am

Coffee & Breakfast

Building 9 - Lobby 08:00 - 08:30 Details

8:30 am

Registration

Building 9 08:30 - 09:00 Details

9:00 am

Welcome Address

  • Takashi Gojobori, KAUST

    Takashi Gojobori

Building 9 - Lecture hall 2 09:00 - 09:05 Details

Takashi Gojobori, KAUST
9:05 am

Opening Remarks

Building 9 - Lecture hall 2 09:05 - 09:20 Details

President Tony Chan, KAUST
Session 1 : Opportunities and Challenges for AI in Medicine (Chair - Takashi Gojobori)
9:20 am

KEYNOTE LECTURE: Estimating genetic and environmental parameters from medical data

In the talk I will cover several approaches for estimating disease-specific parameters from national-scale electronic medical records.

Building 9 - Lecture hall 2 09:20 - 10:10 Details

Andrey Rzhetsky, University of Chicago
10:10 am

Next-generation Sequencing and Inherited Disorders in Saudi Arabia

Building 9 - Lecture hall 2 10:10 - 10:45 Details

Ahmed Alfares, National Guard Health Affairs (NGHA)
10:45 am

Coffee Break

Building 9 - Lobby 10:45 - 11:00 Details

11:00 am

Routes to Drug Design via Bioisosterism

Building 9 - Lecture hall 2 11:00 - 11:35 Details

Alya Arabi, Zayed University
11:35 am

Learning gene functions based on expression: Know what you ask for

Recent updates in sequencing technology have made it possible to start looking at pangenomics questions comparing amongst others gene/presence absence variations in individuals or populations. However when doing so one frequently encounters that many genes lack a functional ontology. Many of these “unknown genes” exist in multiple species, but have not been characterized in any.
Typical approaches to predict the function of those genes comprise phylogenetical profiling and/or to analyze their expression behavior across a large set of experiments. The latter analysis had been perfected for microarray data taking into account sample selection to find genes associated with specific processes, but performance of the same method using modern RNASeq data might have lagged behind. In any case the application of adaptive performance tuning of RNASeq data was hindered by computation intensive procedures for RNASeq data analysis. This issues has been addressed by novel pseudomapping approaches such as those implemented in kallisto/salmon allowing individuals to analyze large expression data matrices. We present data showing that the careful selection of training data positively affects the performance of the outcome and show the effect of novel and fast RNASeq analysis pipelines on gene function predictions using biomedical ontologies.

Building 9 - Lecture hall 2 11:35 - 12:10 Details

Bjorn Usadel, RWTH Aachen University
12:10 pm

Lunch Break

Campus Diner 12:10 - 13:40 Details

Session 2 : New Techniques in AI for the Needs of Medicine and Healthcare ( Chair - Stefan Arold)
1:40 pm

KEYNOTE LECTURE: Accurate, precise and reliable predictions from modelling and simulation using high performance computers

I describe one of our current major research activities which aims to produce actionable outcomes, namely the prediction of binding free energies of small molecules to proteins. We seek, rapid, accurate and reproducible predictions equipped with uncertainty quantification. In general, it is hard to reliably predict by simulation the outcome of a given scientific process. Faced with such difficulties, scientists today often seek to evade the problem by appealing to machine learning. I look at the advantages and disadvantages of invoking such data-driven approaches. Finally, I discuss our recent discovery that much of the true structure of chaotic dynamical systems is lost on digital computers due to their use of IEEE floating point arithmetic. I illustrate this finding with reference to the generalised Bernoulli map, perhaps the simplest of chaotic dynamical systems. I discuss the consequences of this discovery, inter alia for the application of machine learning to simulation data in “AI systems”.

  • Peter Coveney, University College London

    Peter Coveney

Building 9 - Lecture hall 2 13:40 - 14:30 Details

Peter Coveney, University College London
2:30 pm

New perspectives in cancer research: from experimental medicine to Singularity

Building 9 - Lecture hall 2 14:30 - 15:05 Details

Jean-Marc Nabholtz, King Saud Medical City
3:05 pm

Coffee Break

Building 9 - Lobby 15:05 - 15:20 Details

3:20 pm

Differentiable construction of models for the Description Logic EL++

  • Robert Hoehndorf , KAUST

    Robert Hoehndorf

Building 9 - Lecture hall 2 15:20 - 15:55 Details

Robert Hoehndorf , KAUST
3:55 pm

Explainable AI for Precision Medicine & Reprogramming of Cells

  • Jesper Tegner, KAUST

    Jesper Tegner

Building 9 - Lecture hall 2 15:55 - 16:30 Details

Jesper Tegner, KAUST
4:30 pm

Student Poster Session and Finger Buffet - Open to All

Building 9 16:30 - 18:00 Details

8:30 am

Coffee & Breakfast

Building 9 - Lobby 08:30 - 09:10 Details

Session 3 : Drug Development and Repurposing (Chair - Jesper Tegner)
9:10 am

KEYNOTE LECTURE: Machine learning, text mining, and AI approaches for drug discovery and repurposing

Building 9 - Lecture hall 2 09:10 - 10:00 Details

Alexander Tropsha, UNC Eshelman School of Pharmacy, University of North Carolina
10:00 am

The Druggable Genome

Building 9 - Lecture hall 2 10:00 - 10:35 Details

Fowzan Alkuraya, King Faisal Specialist Hospital and Research Center
10:35 am

Coffee Break

Building 9 - Lobby 10:35 - 10:50 Details

10:50 am

How Artificial Intelligence is Changing Drug Discovery for Rare Genetic Disorders

Rare genetic disorders are collectively common specifically in Saudi Arabia because of the high rate of consanguinity. There are a quite virtuous progress in development of drugs for rare genetic disorders. However, the biggest challenge to rare genetic diseases drug development by far is the small size of rare disease populations. Patients with these progressive, serious, life-limiting and life-threatening diseases are often geographically dispersed, which can make it difficult to find enough patients able to reach clinical trial sites. Additionally, it can be difficult for medical practitioners to develop expertise in conditions that are seen so rarely. This often results in a consolidation of expertise at a single or few locations that may be challenging for some patients.
The FDA recognized these challenges and started support to the people affected by rare diseases by accelerating, supporting and facilitating the process of getting orphan drug products to market. In 2016, FDA announced the availability of $2 million in research grants to fund natural history studies in rare diseases. The aim is to collect data on how specific rare diseases progress in individuals over time so that knowledge can inform and support product development and approval. This will be the first time the FDA will provide funding through its Orphan Products grants to conduct these types of studies for rare diseases. Artificial intelligence (AI) is the simulation of human intelligence processes by machines, especially computer systems. These processes include learning (the acquisition of information and rules for using the information), reasoning (using rules to reach approximate or definite conclusions) and self-correction. Recently, it is prove its application in drug discovery. Leading biopharmaceutical companies believe in this new approach. Pfizer is using IBM Watson, a system that uses machine learning, to power its search for immuno-oncology drugs. Sanofi has signed a deal to use UK start-up Exscientia’s artificial-intelligence (AI) platform to hunt for metabolic-disease therapies, and Roche subsidiary Genentech is using an AI system from GNS Healthcare in Cambridge, Massachusetts, to help drive the multinational company’s search for cancer treatments. Most sizeable biopharma players have similar collaborations or internal programmes. Developing expertise in AI for rare genetic disorders is a unique goal that will accelerate drug discovery for such disorders. In conclusion: AI is a new paradigm in medicine and it is application in drug discovery for rare genetic disorders is exciting and need to be explored.

Building 9 - Lecture hall 2 10:50 - 11:25 Details

Majid Alfadhel, King Abdullah International Medical Research Centre
11:25 am

Drug repurposing and target identification through text mining

Building 9 - Lecture hall 2 11:25 - 11:50 Details

Magbubah Essack, KAUST
11:50 am

Group Photo with Conference Speaker

Follow instructions to location 11:50 - 12:00 Details

12:00 pm

Lunch Break

Campus Diner 12:00 - 13:20 Details

Session 4 : Panel Discussion on Requirements of Clinicians and AI scientists (Chair - Charlotte Hauser)
1:20 pm

PANEL DISCUSSION

Building 9 13:20 - 15:20 Details

Ahmed Alfares, National Guard Health Affairs (NGHA) Fowzan Alkuraya, King Faisal Specialist Hospital and Research Center Majid Almadi, King Saud University / King Khalid University Hospital Khalid Al Saleh, King Saud University Jean-Marc Nabholtz, King Saud Medical City Faisal Bindail, AJA Pharma and National Committee for Pharmaceutical Industries Pierre Magistretti, KAUST Takashi Gojobori, KAUST Vladimir Bajic, KAUST
7:00 pm

Gala Dinner - Invitation Only

Yacht Club restaurant 19:00 - 20:30 Details

8:30 am

Coffee & Breakfast

Building 9 - Lobby 08:30 - 09:00 Details

Session 5 : AI and Knowledge Mining from Text, Clinical Data, and Bio-imaging (Chair - Xin Gao)
9:00 am

KEYNOTE LECTURE: Machine Learning for Decision Making in Healthcare

An overview of our ongoing projects aimed to facilitate predictive analytics in healthcare will be presented in this talk. Challenges and the proposed solutions will be discussed related to structured regression on multilayer networks, recovering network connectivity, modeling positive and negative influences, uncertainty propagation and effective integration of domain knowledge and big data. The algorithms will be evaluated in the context of applications related to exploiting information extracted from electronic health records for identifying resources a patient would need for triage systems in emergency departments, estimating hospitalization cost, predicting admission and mortality rate for high impact diseases, identifying disease relationships, discovering gene-disease interactions and assessing tolerance to viral infections.

  • Zoran Obradovic, Temple Univeristy

    Zoran Obradovic

Building 9 - Lecture hall 2 09:00 - 09:50 Details

Zoran Obradovic, Temple Univeristy
9:50 am

Mining Scholarly Data: from Search to Recommendation

: The scientific publication outputs have been exponentially growing in all domains, making document retrieval like finding a needle in a haystack. The main challenge roots in the fast-expanding attributed network with rich information of paper contents on nodes and citation relations on edges. This talk will introduce deep neural network models and deep generative models that are designed for addressing the problems of recommendation of top-k relevant papers to read/cite; identification of top-k possible authors of an anonymous paper; search of datasets and related papers, and identification top-k popular datasets in different domains.

  • Xiangliang Zhang, KAUST

    Xiangliang Zhang

Building 9 - Lecture hall 2 09:50 - 10:10 Details

Xiangliang Zhang, KAUST
10:10 am

Signal processing based features generation for the classification of biomedical signals

  • Meriem T. Laleg, KAUST

    Meriem T. Laleg

Building 9 - Lecture hall 2 10:10 - 10:45 Details

Meriem T. Laleg, KAUST
10:45 am

Coffee Break

Building 9 - Lobby 10:45 - 11:00 Details

11:00 am

Use of ontologies and text mining in biomedicine: applications to infectious disease research

The volume of literature data increases tremendously which makes it inevitable to apply text mining methods to retrieve, manage and analyse the information. Ontologies represent information semantically by describing the knowledge as a set of concepts within specific domains and the relationships between these concepts. Text mining profits from biomedical ontologies in different tasks, especially in mapping ontology concepts to terms in text (e.g. named entity recognition) since they provide domain knowledge. Ontology development profits from automated text mining methods since publications contain references to existing terms and concepts.

In this talk, I will discuss different approaches in which ontologies and text-mining methods have been used in biomedicine with a special focus on infectious diseases. More specifically, I will present on the development of PathoPhenoDB - a database containing pathogen-disease phenotype relations for supporting infectious disease research; use of text mining for expanding the Disease Ontology and use of ontologies and text mined data for predicting host-pathogen interactions.

Building 9 - Lecture hall 2 11:00 - 11:35 Details

Senay Kafkas, KAUST
11:35 am

Mining Biomedical Entity Associations from Text, a DES Perspective

Building 9 - Lecture hall 2 11:35 - 11:55 Details

Adil Salhi
12:00 pm

Lunch Break

Campus Diner 12:00 - 13:15 Details

Session 6 : Spotlight on Young Talent ( Chair - John Archer)
1:15 pm

Predicting multiple toxicity endpoints: Application of multi-label classification with missing labels

Identifying toxicity effects of chemicals is a necessary step in many processes including drug design. To identify associations between chemicals and toxicity effects, we applied multi-label classification (MLC) methods. These methods have not undergone comprehensive benchmarking in the domain of predictive toxicology that could help in identifying guidelines for overcoming the existing deficiencies of these methods. Therefore, we performed extensive benchmarking and analysis of ~19,000 MLC models. We demonstrated variability in the performance of these models under several conditions and determined the best performing model that achieves accuracy of 91% on an independent testing set.

  • Arwa A. Bin Raies

    Arwa A. Bin Raies

Building 9 - Lecture hall 2 13:15 - 13:30 Details

Arwa A. Bin Raies
1:30 pm

Automatic identification of small molecules that promote cell conversion

Cell reprogramming has enormous potential for regenerative medicine and drug discovery. Small molecules have recently been reported to either induce or enhance reprogramming. Exhaustive screens of small molecules are expensive and time consuming given the structural and functional diversity of small molecules and the unlimited number of combinations. We developed a method for the identification of small compounds that either alone or in combinations facilitate the efficacy of cell conversion. Based on comparing primary cell type expression profiles to publicly available drug response expression profiles, the method is able to accurately predict single drugs or drug pairs that can drive any source cell type towards the desired lineage.

Building 9 - Lecture hall 2 13:30 - 13:45 Details

Trisevgeni Rapakoulia, CBRC
1:45 pm

Microbiologically-Influenced Corrosion in oil plant pipelines: the comparative metagenomic approach for the resolution with functional annotation by possible use of AI

The corrosion of ferrous metals in the crude oil industry costs hundreds of millions of dollars annually. It was estimated that about 15% of metal corrosion is due to bacterial activities in a process known as microbiologically-influenced corrosion (MIC). The aim of this study is to develop the method to identify microbial organisms that play key roles for MIC by an analysis of metagenome with bioinformatics.
We obtained scraper samples that were generated in pipe wash operations from seawater injection pipelines and oil pipelines of the oil plants in KSA. We extracted DNA from the samples and generated the 16S rRNA amplicon metagenomes. We also retrieved publically available metagenome data of MIC sites (mainly from USA) from NCBI, using them for the comparison with our metagenomes.

The metagenomes of our scraper samples clearly showed that microflora in seawater were very similar among samples and among oil pipeline samples, respectively. On the other hand, microflora were quite different between seawater and oil pipeline samples. The clustering analysis of our data together with public metagenomes also showed that metagenomes of seawater and oil pipelines were clustered separately. In addition, we found that metagenomes of oil pipeline formed a cluster with metagenomes of MIC sites at deep subsurface environments, showing that microflora in oil pipeline shared the common characteristics of the microbial composition among MICs in anaerobic environments. These results suggest that we should take different strategies against MICs at least between seawater and oil pipelines.

Our results suggest that key microorganisms of MIC are distinct between seawater pipeline and oil pipeline. Because we are producing the Big Data of the MIC metagenomes, we are planning to utilize AI for functional annotation from taxonomic information. Our study leads to the development of a new method to predict, diagnose and provide appropriate treatment for MIC.

  • Badoor Nasser

Building 9 - Lecture hall 2 13:45 - 14:00 Details

Badoor Nasser
2:00 pm

Vec2SPARQL: integrating SPARQL queries and knowledge graph embeddings

Recent developments in machine learning have led to a rise of large number of methods for extracting features from structured data. The features are represented as vectors and may encode for some semantic aspects of data. They can be used in a machine learning models for different tasks or to compute similarities between the entities of the data.
SPARQL is a query language for structured data originally developed for querying Resource Description Framework (RDF) data. It has been in use for over a decade as a standardized NoSQL query language. Many different tools have been developed to enable data sharing with SPARQL. For example, SPARQL endpoints make your data interoperable and available to the world. SPARQL queries can be executed across multiple endpoints. We have developed a Vec2SPARQL, which is a general framework for integrating structured data and their vector space representations. Vec2SPARQL allows jointly querying vector functions such as computing similarities (cosine, correlations) or classifications with machine learning models within a single SPARQL query. We demonstrate applications of our approach for biomedical and clinical use cases.

Building 9 - Lecture hall 2 14:00 - 14:15 Details

Maxat Kulmanov, KAUST
2:15 pm

Coffee Break

Building 9 - Lobby 14:15 - 14:30 Details

2:30 pm

Focal onset seizure prediction using convolutional networks

This work investigates the hypothesis that focal seizures can be predicted using scalp electroencephalogram (EEG) data. Using deep learning, we extract representations that distinguishes between EEG activity relevant to seizure prediction. Convolutional filters on the wavelet transformation of the EEG signal are used to define and learn quantitative signatures of three types of EEG activity: nonseizure, preseizure, and seizure. The optimal seizure prediction horizon is also learned from the data as opposed to making an a priori assumption. In our results, computational solutions to the optimization problem indicate a ten-minute seizure prediction horizon. This result is verified by measuring Kullback-Leibler divergence on the distributions of the automatically extracted features. The results on the EEG database of 204 recordings demonstrate that (i) the preseizure phase transition occurs approximately ten minutes before seizure onset, and (ii) the prediction results on the test set are promising, with a sensitivity of 87.8\% and a low false prediction rate of 0.142 FP/h. Our results significantly outperform a random predictor and other seizure prediction algorithms. We demonstrate that a robust set of features can be learned from scalp EEG that characterize the preseizure state of focal seizures.

  • Haidar Khan

    Haidar Khan

Building 9 - Lecture hall 2 14:30 - 14:45 Details

Haidar Khan
2:45 pm

Closing Remarks and Announcement of Poster Competition Winners

  • Takashi Gojobori, KAUST

    Takashi Gojobori

Building 9 - Lecture hall 2 14:45 - 15:00 Details

Takashi Gojobori, KAUST
7:00 pm

Dinner - Invitation Only

Golf Club 19:00 - 20:30 Details