Multinode Multi-GPU Two-Electron Integrals: Code Generation Using the Regent Language.

The computation of two-electron repulsion integrals (ERIs) is often the most expensive step of integral-direct self-consistent field methods. Formally it scales as O(N4), where N is the number of Gaussian basis functions used to represent the molecular wave function. In practice, this scaling can be reduced to O(N2) or less by neglecting small integrals with screening methods. The contributions of the ERIs to the Fock matrix are of Coulomb (J) and exchange (K) type and require separate algorithms to compute matrix elements efficiently. We previously implemented highly efficient GPU-accelerated J-matrix and K-matrix algorithms in the electronic structure code TeraChem. Although these implementations supported the use of multiple GPUs on a node, they did not support the use of multiple nodes. This presents a key bottleneck to cutting-edge ab initio simulations of large systems, e.g., excited state dynamics of photoactive proteins. We present our implementation of multinode multi-GPU J- and K-matrix algorithms in TeraChem using the Regent programming language. Regent directly supports distributed computation in a task-based model and can generate code for a variety of architectures, including NVIDIA GPUs. We demonstrate multinode scaling up to 45 GPUs (3 nodes) and benchmark against hand-coded TeraChem integral code. We also outline our metaprogrammed Regent implementation, which enables flexible code generation for integrals of different angular momenta.

Progression to active tuberculosis, but not transmission, varies by Mycobacterium tuberculosis lineage in The Gambia.

BACKGROUND: There is considerable variability in the outcome of Mycobacterium tuberculosis infection. We hypothesized that Mycobacterium africanum was less likely than M. tuberculosis to transmit and progress to tuberculosis disease. METHODS: In a cohort study of patients with tuberculosis and their household contacts in The Gambia, we categorized 1808 HIV-negative tuberculosis contacts according to exposure to M. tuberculosis or M. africanum. Positive skin test results indicated transmission, and development of tuberculosis during 2 years of follow-up indicated progression to disease. RESULTS: Transmission rates were similar, but rates of progression to disease were significantly lower in contacts exposed to M. africanum than in those exposed to M. tuberculosis (1.0% vs. 2.9%; hazard ratio [HR], 3.1 [95% confidence interval {CI}, 1.1-8.7]). Within M. tuberculosis sensu stricto, contacts exposed to a Beijing family strain were most likely to progress to disease (5.6%; HR relative to M. africanum, 6.7 [95% CI, 2.0-22]). CONCLUSIONS: M. africanum and M. tuberculosis transmit equally well to household contacts, but contacts exposed to M. africanum are less likely to progress to tuberculosis disease than those exposed to M. tuberculosis. The variable rate of progression by lineage suggests that tuberculosis variability matters in clinical settings and should be accounted for in studies evaluating tuberculosis vaccines and treatment regimens for latent tuberculosis infection.

Incidence of tuberculosis and the predictive value of ELISPOT and Mantoux tests in Gambian case contacts.

BACKGROUND: Studies of Tuberculosis (TB) case contacts are increasingly being utilised for understanding the relationship between M. tuberculosis and the human host and for assessing new interventions and diagnostic tests. We aimed to identify the incidence rate of new TB cases among TB contacts and to relate this to their initial Mantoux and ELISPOT test results. METHODS AND FINDINGS: After initial Mantoux and ELISPOT tests and exclusion of co-prevalent TB cases, we followed 2348 household contacts of sputum smear positive TB cases. We visited them at 3 months, 6 months, 12 months, 18 months and 24 months, and investigated those with symptoms consistent with TB. Those who were diagnosed separately at a government clinic had a chest x-ray. Twenty six contacts were diagnosed with definite TB over 4312 person years of follow-up (Incidence rate 603/100,000 person years; 95% Confidence Interval, 370-830). Nine index and secondary case pairs had cultured isolates available for genotyping. Of these, 6 pairs were concordant and 3 were discordant. 2.5% of non-progressors were HIV positive compared to 12% of progressors (HR 6.2; 95% CI 1.7-22.5; p = 0.010). 25 secondary cases had initial Mantoux results, 14 (56%) were positive ; 21 had initial ELISPOT results, 11 (52%) were positive; 15 (71%) of 21 tested were positive by one or the other test. Of the 6 contacts who had concordant isolates with their respective index case, 4 (67%) were Mantoux positive at recruitment, 3 (50%) were ELISPOT positive; 5 (83%) were positive by one or other of the two tests. ELISPOT positive contacts, and those with discordant results, had a similar rate of progression to those who were Mantoux positive. Those negative on either or both tests had the lowest rate of progression. CONCLUSIONS: The incidence rate of TB disease in Gambian TB case contacts, after screening for co-prevalent cases, was 603/100,000 person years. Since initial ELISPOT test and Mantoux tests were each positive in only just over half of cases, but 71% were positive by one or other test, positivity by either might be the best indication for preventive treatment. These data do not support the replacement of the Mantoux test by an ELISPOT test in The Gambia or similar settings.

Longitudinal assessment of an ELISPOT test for Mycobacterium tuberculosis infection.

BACKGROUND: Very little longitudinal information is available regarding the performance of T cell-based tests for Mycobacterium tuberculosis infection. To address this deficiency, we conducted a longitudinal assessment of the enzyme-linked immunosorbent spot test (ELISPOT) test in comparison to the standard tuberculin skin test (TST). METHODS AND FINDINGS: In tuberculosis (TB) contacts we repeated ELISPOT tests 3 mo (n = 341) and 18 mo (n = 210) after recruitment and TSTs at 18 mo (n = 130). We evaluated factors for association with conversion and reversion and investigated suspected cases of TB. Of 207 ELISPOT-negative contacts, 51 (24.6%) had 3-mo ELISPOT conversion, which was associated with a positive recruitment TST (odds ratio [OR] 2.2, 95% confidence interval [CI] 1.0-5.0, p = 0.048) and negatively associated with bacillus Calmette-Guérin (BCG) vaccination (OR 0.5, 95% CI 0.2-1.0, p = 0.06). Of 134 contacts, 54 (40.2%) underwent 3-mo ELISPOT reversion, which was less likely in those with a positive recruitment TST (OR 0.3, 95% CI 0.1-0.8, p = 0.014). Between 3 and 18 mo, 35/132 (26.5%) contacts underwent ELISPOT conversion and 28/78 (35.9%) underwent ELISPOT reversion. Of the 210 contacts with complete results, 73 (34.8%) were ELISPOT negative at all three time points; 36 (17.1%) were positive at all three time points. Between recruitment and 18 mo, 20 (27%) contacts had ELISPOT conversion; 37 (50%) had TST conversion, which was associated with a positive recruitment ELISPOT (OR 7.2, 95% CI 1.4-37.1, p = 0.019); 18 (32.7%) underwent ELISPOT reversion; and five (8.9%) underwent TST reversion. Results in 13 contacts diagnosed as having TB were mixed, but suggested higher TST sensitivity. CONCLUSIONS: Both ELISPOT conversion and reversion occur after M. tuberculosis exposure. Rapid ELISPOT reversion may reflect M. tuberculosis clearance or transition into dormancy and may contribute to the relatively low reported ELISPOT conversion rate. Therefore, a negative ELISPOT test for M. tuberculosis infection should be interpreted with caution.