JP's Notes

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Computational Life Sciences ======

{{ :images:umlogo_redrgb.png?nolink&200|}}This page lists all the talks/workshops/activities of the Computational Life Sciences (cls-uom) research group at the University of Malta. 

This group is interested in applying computer science techniques to problems in molecular biology, chemistry, pharmacology, and drug-discovery.  As biomedical and healthcare research becomes more interdisciplinary, we realize the need to bring computer scientists, wet-lab biologists, mathematicians, statisticians, geneticists, //etc.// together.  We organize talks and workshops, mostly in bioinformatics and cheminformatics, where academics and post-graduate students present their work in an informal and friendly setting.

For more information, and to receive notifications of future activities please subscribe to our [[https://groups.google.com/d/forum/cls-um|Google Group]].

===== Talks =====

The following is a (reverse) chronological ordered list of talks.

^ Date/Time ^ Speaker     ^ Location ^ Title & Abstract ^
|  Fri 24th May 2019, 12:00pm-1:00pm | Dr Jean-Paul Ebejer | Faculty of ICT, Networks Lab, Level -1 block B, Room 7 | **Machine Learning in Computer-Aided Drug Discovery (Workshop)** \\ Computer-Aided Drug Design (CADD) plays an increasingly critical role in the drug-discovery process.  CADD involves the application of computer algorithms to improve pharmaceutical productivity.  These include algorithms for the identification of the biological target involved in a disease, toxicity and side-effect prediction, and searching a database for molecules which exhibit a therapeutic effect against a particular protein of interest.  The latter is known as Virtual Screening. In this workshop I will give an overview of CADD with particular emphasis on virtual screening.  We will develop a machine learning (ML) model to discriminate between actives and decoys against a protein target which plays a critical role in the life-cycle of HIV.  This interdisciplinary talk is aimed at an audience with prior Python programming experience and an interest in the application of ML models in life sciences. This talk is part of the events organised by the [[https://www.um.edu.mt/research/researchprojects/dsrg|Data-Science Research Platform]]. |
|  Fri 15th March 2019, 2:30-3:30pm | Mr Kenneth Penza | Faculty of ICT, Room 9, Level -1, Block A | **GO term predictions in CATH: a machine learning approach** \\ Proteins are composed of amino acids chains that perform tasks such as regulation and signalling within an organism. The protein folds into a three-dimensional structure, giving it functionality. A protein function is characterised through laboratory experiments or computational methods. Physical laboratory experiments are expensive but more reliable with a low throughput. Two proteins are related if they have a common ancestor. The ancestry (homology) link is established through conserved regions in the protein sequence. Homology relationships can be utilised to transfer protein functions between related proteins. This research investigates machine learning techniques to predict protein function from protein properties and proportions from structural databases such as CATH and PFAM. This work used two Machine Learning (ML) approaches to tackle this problem. The first approach was the application of automatic feature selection using Support Vector Machines (SVM) and Random Forest (RF) techniques on species-specific datasets. The second approach applied the different species-specific datasets to Neural Networks with different hidden layer configurations. These techniques were evaluated on CAFA3 targets against the CAFA2 shared task. The RF models with the species-specific feature set performs at the same level of the best CAFA2 submission for Homo sapiens species and is superior to the best CAFA2 submission for E. coli.|
|  Wed 16th January 2019, 12-1pm | Mr Nicholas Mamo | BM402 (CMMB, Biomedical Sciences Building) | **Demystifying Blockchain Technology: The Blockchain for Non-Computer Scientists** \\ If you live in Malta, then you also inhabit the self-proclaimed Blockchain Island. What does that actually mean? What is the blockchain, and how will it change the way that we think about data?  Blockchains are mentioned in the same breath as crypto-currencies, but their applicability extends beyond the world of finance. This talk breaks down the technology in non-technical terms to debunk its misconceptions and understand how it could impact healthcare. We will explain the main properties of blockchain systems, and describe requirements necessary for the correct application of blockchain technologies.  Nicholas Mamo read an undergraduate degree in AI and is presently a Research Support Officer at the Centre for Molecular Medicine and Biobanking working on the externally-funded “Connecting for Health” project.|
|  Tue 10th April 2018, 4-5pm | Mr Karl Pullicino | ICT Faculty Building, CS seminar room 38, Block B, 1st floor | **A MapReduce approach to Genome Alignment** \\ Recent years brought an enormous growth in DNA sequencing capacity and speed, thanks to the application of next-generation sequencing (NGS) technologies. The alignment of read sequences to a given reference genome is crucial for further diagnostic downstream analysis. Finding the optimal alignment of short DNA reads from a biological sample to a reference human genome, requires big data techniques, since reads’ size are in the region of 200GB. In this dissertation we present three approaches to perform distributed sequence alignment of genomic data. The first one is based on an optimization of the Smith-Waterman algorithm. The other two approaches are based on the MapReduce programming paradigm. MR-BWA presents a novel approach in distributing BWA in a different manner than existing work. BWA is an industry standard software used for genomic reads alignment. MR-BWT-FM presents low level optimizations on suffix array and BWT creation which are used to create a custom FM-Index which in turn is used for distributed genome sequence alignment. Output generated by the application generates insights and charts about the results. We evaluate the performance and correctness of both approaches by comparing our output with that of similar tools, using standard datasets from the 1000 Genomes Project. Performance and correctness results for both distributed approaches are comparable with similar tools, whilst the final custom FM-Index size is smaller than the standard BWA index size. The source code of the software described in this dissertation is publicly available at {{https://github.com/kpullu/msc|https://github.com/kpullu/msc}}.|
| Mon 17th July 2017, 4pm | Mr Joseph Bonello | ICT Faculty Building, CS seminar room 38, Block B, 1st floor | **Protein Function Prediction Using Homologues** {{ :201707_cls_porteinfunctionpredicitionusinghomologs_mrjosephbonello.pdf |Slides}} \\ Homology refers to the existence of a common origin between a pair of proteins in different organisms. Proteins consist of multiple domains – conserved regions of a sequence and structure that can function independently from the rest of the protein chain. Protein function prediction methods based on homology, take advantage of the many pairwise homology relationships between individual domain sequences. \\ This project attempts to create a set of scores that can be used to predict the possible domain functions that a protein can possess. The study uses CATH Superfamilies and CATH Functional Families (FunFams) to generate the scores. CATH is a database that provides a hierarchical protein-domain classification for proteins obtained from PDB. The Superfamilies and FunFams provide a natural grouping for proteins that share the same evolutionary origin (homologous superfamilies). This grouping can be exploited to generate similarity scores between the domains and the families.  Two methods have been developed for the purpose of function prediction based on these principles. The first method uses Set Theory, where the proteins belonging to a Superfamily or a FunFam are used to determine which GO Terms are more likely to occur in the group. The second method uses a statistical calculation to represent the presence of GO Terms in a family.|
| Thu 1st June 2017, 4pm | Dr Jean-Paul Ebejer | ICT Faculty Building, CS seminar room 38, Block B, 1st floor | **Computer-Aided Drug Design**  \\ Computer-Aided Drug Design (CADD) plays an increasingly critical role in the drug-discovery process. CADD involves the application of computer algorithms to improve pharmaceutical productivity. These include algorithms for the identification of the biological target involved in a disease, toxicity and side-effect prediction, and searching a database for molecules which exhibit a therapeutic effect against a particular protein of interest. The latter is known as Virtual Screening. In this talk I will give an overview of CADD with particular emphasis on virtual screening. I will describe the successes, challenges and limitations of the approach. Finally, I will briefly present a novel virtual screening method we have developed. This interdisciplinary talk is aimed at an audience of broad interest. |