Search for: All records

Residue Number Systems (RNS) demonstrate the fascinating potential to serve integer addition/multiplication-intensive applications. The complexity of Artificial Intelligence (AI) models has grown enormously in recent years. From a computer system’s perspective, ensuring the training of these large-scale AI models within an adequate time and energy consumption has become a big concern. Matrix multiplication is a dominant subroutine in many prevailing AI models, with an addition/multiplication-intensive attribute. However, the data type of matrix multiplication within machine learning training typically requires real numbers, which indicates that RNS benefits for integer applications cannot be directly gained by AI training. The state-of-the-art RNS real number encodings, including floating-point and fixed-point, have defects and can be further enhanced. To transform default RNS benefits to the efficiency of large-scale AI training, we propose a low-cost and high-accuracy RNS fixed-point representation: Single RNS Logical Partition (S-RNS-Logic-P) representation with Scaling Down Postprocessing Multiplication (SD-Post-Mul). Moreover, we extend the implementation details of the other two RNS fixed-point methods: Double RNS Concatenation (D-RNS-Concat) and Single RNS Logical Partition (S-RNS-Logic-P) representation with Scaling Down Preprocessing Multiplication (SD-Pre-Mul). We also design the architectures of these three fixed-point multipliers. In empirical experiments, our S-RNS-Logic-P representation with SD-Post-Mul method achieves less latency and energy overhead while maintaining good accuracy. Furthermore, this method can easily extend to the Redundant Residue Number System (RRNS) to raise the efficiency of error-tolerant domains, such as improving the error correction efficiency of quantum computing. 

Severe Acute respiratory syndrome coronavirus (SARS-CoV-1) attaches to the host cell surface to initiate the interaction between the receptor-binding domain (RBD) of its spike glycoprotein (S) and the human Angiotensin-converting enzyme (hACE2) receptor. SARS-CoV-1 mutates frequently because of its RNA genome, which challenges the antiviral development. Here, we per-formed computational saturation mutagenesis of the S protein of SARS-CoV-1 to identify the residues crucial for its functions. We used the structure-based energy calculations to analyze the effects of the missense mutations on the SARS-CoV-1 S stability and the binding affinity with hACE2. The sequence and structure alignment showed similarities between the S proteins of SARS-CoV-1 and SARS-CoV-2. Interestingly, we found that target mutations of S protein amino acids generate similar effects on their stabilities between SARS-CoV-1 and SARS-CoV-2. For example, G839W of SARS-CoV-1 corresponds to G857W of SARS-CoV-2, which decrease the stability of their S glycoproteins. The viral mutation analysis of the two different SARS-CoV-1 isolates showed that mutations, T487S and L472P, weakened the S-hACE2 binding of the 2003–2004 SARS-CoV-1 isolate. In addition, the mutations of L472P and F360S destabilized the 2003–2004 viral isolate. We further predicted that many mutations on N-linked glycosylation sites would increase the stability of the S glycoprotein. Our results can be of therapeutic importance in the design of antivirals or vaccines against SARS-CoV-1 and SARS-CoV-2. 

• The structure-based energy calculations were applied to determine the effects of disease-causing kinesin missense mutations on protein stability and protein-protein interaction. • The mutations associated with Intellectual Disability can decrease the protein stability of KIF1A motor domain. • Hereditary Spastic Paraplegia mutations located in kinesin-tubulin complex interface can destabilize the binding infinity of KIF5A-tubulin complex.