Clustering scientific documents with topic modeling

Topic modeling is a type of statistical model for discovering the latent “topics” that occur in a collection of documents through machine learning. Currently, latent Dirichlet allocation (LDA) is a popular and common modeling approach. In this paper, we investigate methods, including LDA and its extensions, for separating a set of scientific publications into several clusters. To evaluate the results, we generate a collection of documents that contain academic papers from several different fields and see whether papers in the same field will be clustered together. We explore potential scientometric applications of such text analysis capabilities.     

Keyword guessing attacks on secure searchable public key encryption schemes with a designated tester

The first searchable public key encryption scheme with designated testers (dPEKS) known to be secure against keyword guessing attacks was due to Rhee et al. [H.S. Rhee, W. Susilo, and H.J. Kim, Secure searchable public key encryption scheme against keyword guessing attacks, IEICE Electron. Express 6(5) (2009), pp. 237–243]. Recently, some dPEKS schemes, including the Rhee et al. scheme, were found to be vulnerable to keyword guessing attacks by a malicious server. However, the Rhee et al. dPEKS scheme and its improved variants are still known to be secure against keyword guessing attack by the outsider attacker to date. In this paper, we present a keyword guessing attack by the outsider attacker on the existing dPEKS schemes. We first describe the attack scenario which is possible in the current nature of the Internet and public key encryption with keyword search applications, e.g. email routing. We then demonstrate the detailed attack steps on the Rhee et al. scheme as an attack instance. We emphasize that our attack is generic and it equally applies to all existing dPEKS schemes that claim to be secure against keyword guessing attacks by the outsider attacker.     

Residue-specific mass signatures for the efficient detection of protein modifications by mass spectrometry

Currently available mass spectrometric (MS) techniques lack specificity in identifying protein modifications because molecular mass is the only parameter used to characterize these changes. Consequently, the suspected modified peptides are subjected to tandem MS/MS sequencing that may demand more time and sample. We report the use of stable isotope-enriched amino acids as residue-specific "mass signatures" for the rapid and sensitive detection of protein modifications directly from the peptide mass map (PMM) without enrichment of the modified peptides. These mass signatures are easily recognized through their characteristic spectral patterns and provide fingerprints for peptides containing the same content of specific amino acid residue(s) in a PMM. Without the need for tandem MS/MS sequencing, a peptide and its modified form(s) can readily be identified through their identical fingerprints, regardless of the nature of modifications. In this report, we demonstrate this strategy for the detection of methionine oxidation and protein phosphorylation. More interestingly, the phosphorylation of a histone protein, H2A.X, obtained from human skin fibroblast cells, was effectively identified in response to low-dose radiation. In general, this strategy of residue-specific mass tagging should be applicable to other posttranslational modifications.