Shereena M. Arif

Analysis and use of fragment-occurrence data in similarity-based virtual screening

Current systems for similarity-based virtual screening use similarity measures in which all the fragments in a fingerprint contribute equally to the calculation of structural similarity. This paper discusses the weighting of fragments on the basis of their frequencies of occurrence in molecules. Extensive experiments with sets of active molecules from the MDL Drug Data Report and the World of Molecular Bioactivity databases, using fingerprints encoding Tripos holograms, Pipeline Pilot ECFC_4 circular substructures and Sunset Molecular keys, demonstrate clearly that frequency-based screening is generally more effective than conventional, unweighted screening. The results suggest that standardising the raw occurrence frequencies by taking the square root of the frequencies will maximise the effectiveness of virtual screening. An upper-bound analysis shows the complex interactions that can take place between representations, weighting schemes and similarity coefficients when similarity measures are computed, and provides a rationalisation of the relative performance of the various weighting schemes.

Reconstructing gene regulatory networks from knock-out data using Gaussian Noise Model and Pearson Correlation Coefficient

A gene regulatory network (GRN) is a large and complex network consisting of interacting elements that, over time, affect each others state. The dynamics of complex gene regulatory processes are difficult to understand using intuitive approaches alone. To overcome this problem, we propose an algorithm for inferring the regulatory interactions from knock-out data using a Gaussian model combines with Pearson Correlation Coefficient (PCC). There are several problems relating to GRN construction that have been outlined in this paper. We demonstrated the ability of our proposed method to (1) predict the presence of regulatory interactions between genes, (2) their directionality and (3) their states (activation or suppression). The algorithm was applied to network sizes of 10 and 50 genes from DREAM3 datasets and network sizes of 10 from DREAM4 datasets. The predicted networks were evaluated based on AUROC and AUPR. We discovered that high false positive values were generated by our GRN prediction methods because the indirect regulations have been wrongly predicted as true relationships. We achieved satisfactory results as the majority of sub-networks achieved AUROC values above 0.5.

Enhancing the Effectiveness of Fingerprint-Based Virtual Screening: Use of Turbo Similarity Searching and of Fragment Frequencies of Occurrence

Binary fingerprints encoding the presence of 2D fragment substructures in molecules are extensively used for similarity-based virtual screening in the agrochemical and pharmaceutical industries. This paper describes two techniques for enhancing the effectiveness of screening: the use of a second-level search based on the nearest neighbours of the initial reference structure; and the use of weighted fingerprints encoding the frequency of occurrence, rather than just the mere presence, of substructures. Experiments using several databases for which both structural and bioactivity data are available demonstrate the effectiveness of these two approaches.

Comparison of chemical similarity measures using different numbers of query structures

Many different similarity measures have been described for searching chemical databases. Drawing on previous work in textual information retrieval, this paper investigates the numbers of queries that are required to make robust statements as to the relative retrieval effectiveness of different similarity measures. Experiments with the MDL Drug Data Report database suggest that much larger numbers of queries are ideally required for this purpose than has been the case in previous comparative studies in chemoinformatics.

The Use of Weighted 2D Fingerprints in Similarity-Based Virtual Screening

Abstract The fingerprints that are widely used for similarity-based virtual screening typically encode the presence or absence of fragments, without any indication as to their relative importance. This chapter discusses the use of weighted fingerprints, where each fragment is associated with a weight denoting its degree of importance in quantifying the degree of similarity between a reference structure and a database structure. Extensive studies using the World of Molecular Bioactivity and MDL Drug Data Report databases show that weighting fragments according to their frequency of occurrence within a molecule can increase the effectiveness of screening, but that this is not the case when fragments are weighted according to their frequency of occurrence within a database.

New strategy for Turbo Similarity Searching: Implementation and testing

Virtual screening is one of the most vital methods applied in Chemoinformatics, the field that contributes to drug discovery process. Turbo Similarity Searching (TSS) and data fusion are two of the latest chemical similarity searching strategies, which has evolved from the conventional similarity searching (SS) that apply the concept of multi-target searching instead of just an individual target search. The indirect relationship exists in TSS, with the inclusion of Nearest Neighbours (NN) has been proven to have better performance than the direct relationship (i.e. between query structure and database structures) that exists in similarity searching process. In this paper, we will focus on the implementation and improvement of the existing TSS. By adding in another layer of indirect relationship between the reference compound and the database compounds, along with an additional fusion layer, the performance of the new TSS strategy can be observed. The initial results indicated that there is an obvious increment in the recall value when applying the new strategy. The results are also evaluated with the significance test to show that the result produced by the new strategy is true and does not occurred by chance. Further work on different activity classes and different descriptors on the new strategy are expected to generate a better performance than the existing TSS.

The Effect of Noise Elimination and Stemming in Sentiment Analysis for Malay Documents

The growth of technology has changed the way of communicating opinions on services and products. In consumerism, the real challenge is to understand the latest trends and summarize the state or general opinions about products due to the diversity and size of social media data such as Twitter, Facebook and online forum. This paper discusses sentiments analysis in Malay documents from three perspectives. First, several alternatives of text representation were investigated. Second, the effects of the pre-processing strategies such as normalization and stemming with two type of Malay stemmer algorithm were highlighted. And lastly, the performance of Naive Bayes (NB), Support Vector Machine (SVM) and K-nearest neighbour (kNN) classifiers in classifying positive and negative reviews, were compared. The results show that our selection of pre-processing strategies on the reviews slightly increases the performance of the classifiers.

Comparison of Similarity Coefficients for Chemical Database Retrieval

Similarity-based virtual screening is used in drug discovery by using computational model for rapid evaluation of large number of chemical molecules. Similarity searches use 2D or 3D fingerprints and similarity coefficient to calculate the structural resemblance between each molecule in a chemical database and a target structure. The objective of this work is to determine the best coefficient to be used in similarity searching to get the optimal results. This paper will describe the experiment to perform the molecular similarity searching using different similarity coefficients, which focus on 2D UNITY or ECFP4 fingerprint on 5 activity classes. We will also highlight the different similarity values and the optimal results of similarity measures. All this could depend on what type of fingerprint. As a conclusion, we found that every combination measure has its own advantage. But to look for the best possible results, the nature of molecular activity class could also play an important role.

Defeasible Policy Language for Online Social Networks

Current online social network sites have not addressed policy control sufficiently. In addition, the existing rule-based proposals on policy control have not been able to cope with normative, temporal, exceptional and conflicting nature of OSNs policies. These characteristics of OSNs policies fit to defeasible logic formalism. Thus, we contextualized a defeasible policy language and proposed corresponding ontologies to extend an existing ontology framework on policy control called open digital right language. Our ontology proposal focused on OSNs use cases and provided solution for implementing norms, deadlines and conflict resolution and compensation models for policies. Deployed ontology shows that defeasible policy language is expressive enough to represent and manipulate complicated policy use cases of OSNs.

The Effect of Adding Indirect Relationship to Turbo Similarity Searching

Background: Turbo Similarity Searching (TSS) has been proved as one of the effective and simple searching method in Cheminformatics. Emerging from the conventional similarity searching, TSS depended on the concept of fusion where relationship between the target being sought and the compound in the database are indirect. Previous works has looked at only one level of indirect relationship and indicates that there are further potential that more levels of such relationship be added to TSS to increase its ability to recover more actives. Hence, in this work, we aimed to investigate the impact of the indirect relationship on TSS. Method: This study has further investigated the enhancement of TSS using additional layers of indirect relationship and fusion process. We implemented TSS by adding another layer of fusion between the target and database compound. Results: The experiments with MDDR database showed that the proposed new strategy described in this paper provide a way of enhancing the effectiveness of the TSS process in chemical databases. The experiments also showed that the increases in performance are particularly better when the sought actives are structurally diverse. Conclusion: We may conclude that the additional layers do increase the recall of TSS. Hence, the new TSS strategy could be used as an alternative to the old TSS.