MOTIVATION: The extraordinary genetic and antigenic variability of RNA viruses is arguably the greatest challenge to the development of broadly effective vaccines. No single viral variant can induce sufficiently broad immunity, and incorporating all known naturally circulating variants into one multivalent vaccine is not feasible. Furthermore, no objective strategies currently exist to select actual viral variants that should be included or excluded in polyvalent vaccines. RESULTS: To address this problem, we demonstrate a method based on graph theory that quantifies the relative importance of viral variants. We demonstrate our method through application to the envelope glycoprotein gene of a particularly diverse RNA virus of pigs: porcine reproductive and respiratory syndrome virus (PRRSV). Using distance matrices derived from sequence nucleotide difference, amino acid difference and evolutionary distance, we constructed viral networks and used common network statistics to assign each sequence an objective ranking of relative 'importance'. To validate our approach, we use an independent published algorithm to score our top-ranked wild-type variants for coverage of putative T-cell epitopes across the 9383 sequences in our dataset. Top-ranked viruses achieve significantly higher coverage than low-ranked viruses, and top-ranked viruses achieve nearly equal coverage as a synthetic mosaic protein constructed in silico from the same set of 9383 sequences. CONCLUSION: Our approach relies on the network structure of PRRSV but applies to any diverse RNA virus because it identifies subsets of viral variants that are most important to overall viral diversity. We suggest that this method, through the objective quantification of variant importance, provides criteria for choosing viral variants for further characterization, diagnostics, surveillance and ultimately polyvalent vaccine development.

Embedded DRAM offers major advantages in terms of system integration on-chip. However, the fabrication process requires special attention for profitable manufacturing, particularly in the area of test and repair. We present the benefits which embedded DRAM offers, and then move to test issues. We provide an analysis of the technical and financial impact of different test approaches, and discuss which considerations are important in choosing a test strategy.

For pt. 1 see ibid., p. 163-172 (2001). The era of embedded DRAM (eDRAM) in systems-on-silicon and in multimillion gate ASICs is upon us. As process developers and logic designers succeed in offering highly integrated functions to their customers, the test engineer is left with the seemingly impossible task of guaranteeing that the eDRAM utilized in these advanced chips has the same high test coverage and outgoing quality level as the customer has come to expect with discrete DRAM. Unfortunately, some of the very attributes that make eDRAM attractive to the ASIC designer i.e., high bandwidth with high I/O bus width and thus high macro pin count, make it very difficult to test when isolated from the chip terminals by various levels of logic and instantiated multiple times in the hierarchy. Several differing approaches have been taken to attempt to optimize the test of the embedded memory in these complex chips. This paper investigates two such approaches that have been developed: the ASIC-centric approach and the semi-custom, high volume solution, and discuss some details of their differing methodology and motivations for their varying directions.

Process scaling and the need for smaller SRAM cells challenges process technologies to make millions of robust and reliable bitline contacts on a single chip. Another challenge is to identify marginal, resistive and unreliable bitline contacts given the inherent electrical characteristics of the SRAM cell. This paper describes two design techniques that improve the screenability and stressability of bitline contacts in high-density SRAMs. These techniques are developed to overcome the lack of detectability of resistive bitline contacts in SRAM cells.