Session S3

Session S3: Genomics and Signal Processing

Time:                 Tuesday, May 13, 10:30-12:00
Chair:                 Zoe Lacroix, Arizona State University
Co-Chair:           Lina Karam, Arizona State University

S3-1: A Low-complexity Probabilistic Genome Assembly Based on Hashing Functions
Naji Mounsef, Lina Karam, Zoe Lacroix and Christophe Legendre Arizona State University, Tempe AZ
This paper presents an efficient and highly accurate low-complexity genome assembly algorithm. The proposed algorithm uses a hashing function to reduce the complexity of the assembly process. The algorithm is tested against datasets of genomic sequences of different lengths. Its performance is evaluated against Phrap, a well-known sequence assembly tool, in terms of completeness, accuracy, and efficiency (time and space). Results show that the proposed assembly algorithm outperforms Phrap in terms of accuracy, time and memory.

S3-2: Protein Secondary Structure Estimation Using Linear Prediction and Cepstral Features
Shibani Misra and Andreas Spanias Arizona State University, Tempe AZ
A method of protein secondary structural classification is proposed that uses the mean of the linear prediction based cepstral feature vectors. The cepstral vectors are extracted from primary protein sequences that are mapped using two existing indexing techniques. The Mahalanobis distance metric is used as it can model the short range correlations between the amino acid sequences better than the Euclidean distance. The performance of the algorithm is evaluated using the resubstitution test, the jackknife test, and the 10-way CV method. The proposed algorithm shows a 3% improvement (both for jackknife and resubstitution) over the autocorrelation function based approach, and an 8% (jackknife) and 6% (resubstitution) improvement over the component-coupled algorithm, an amino acid composition based approach. The robustness of the proposed classifier to noisy data is tested by introducing upto 5 bit errors. The accuracy of the classifier decreases by 10% for more than 3 bit errors.

S3-3: Waveform Mapping based Alignment methods for DNA Sequences
Lakshminarayan Ravichandran, Antonia Papandreou-Suppappola, Andreas Spanias, Zoe Lacroix and Christophe Legendre Arizona State University, Tempe AZ
DNA sequences have been efficiently processed using signal analysis techniques. In order to further explore the use of signal processing in sequencing, we investigate three waveform mapping schemes that map DNA sequence elements to different waveform basis. One of the proposed mapping schemes uses regions in the time-frequency plane to compactly represent the elements and their positioning in a sequence. Sequence alignment is a method that compares and classifies regions of similarity in sequences based on their functional properties. Common alignment techniques like the matched filter approach use cross-correlations that result in many misalignments when capturing local and repetitive alignments. We propose a time-frequency based alignment technique using the matching pursuit decomposition principle and the time-frequency waveform mapping. The aim of this alignment algorithm is to identify local and global alignments efficiently and with high precision.

S3-4: Genomic Processing for Diabetes Mellitus Classification
Raja Rajeswari Konduri Andhra University, India, Pathimati Srihari Dadi Institute of Engineering and Technology, India, Apparao Allam and Sreedhar G.R. Andhra University, India
Genomic signal processing has been defined as the analysis, processing, and use of genomic signals for gaining biological knowledge and the translation of that knowledge into systems-based applications. Diabetes Mellitus is an illness characterized by high levels of circulating glucose in the blood that result from either too little insulin or inadequate insulin functioning. In this paper, the K-means algorithm is applied for clustering genes responsible for diabetes mellitus.