Figure 1. A: The structure of Thermonuclease from Staphylococcus aureus was predicted by AlphaFold2. The change in stability upon mutation, ∆∆G (positive values indicate destabilization), is plotted as a function of effective strain measured at the mutated site, and in a spherical region near the mutated site, for 491 mutants. The black line represents the median. Statistical correlations are displayed for Pearson’s r and Spearman’s ρ.
Figure 1. A: The structure of Thermonuclease from Staphylococcus aureus was predicted by AlphaFold2. The change in stability upon mutation, ∆∆G (positive values indicate destabilization), is plotted as a function of effective strain measured at the mutated site, and in a spherical region near the mutated site, for 491 mutants. The black line represents the median. Statistical correlations are displayed for Pearson’s r and Spearman’s ρ.
Tyler O'Neal, Staff Editor SCIENCE

Research on protein stability at the Institute for Basic Science raises questions

The recent research from the Center for Algorithmic and Robotized Synthesis at the Institute for Basic Science (IBS) in South Korea has generated skepticism within the scientific community. The study suggests that AI-predicted structures, using DeepMind’s AlphaFold2 algorithm, can indicate protein mutant stability. Despite the potentially beneficial implications of the research, many important questions have been raised about the validity and applicability of the findings.

One primary concern is whether AlphaFold2 has genuinely understood the physics of protein folding, or if it functions more as a sophisticated regression model that identifies statistical patterns. The study’s authors, John McBride, and Tsvi Tlusty, sought to assess AlphaFold2's capability to predict the effects of mutations on stability, a challenging task given the multitude of potential mutations and the intricate nature of protein folding.

The research results propose that the structural changes anticipated by AlphaFold2 are connected to changes in stability induced by mutations. While this discovery could imply that the algorithm contains valuable stability information, critics have raised doubts about the depth and accuracy of these predictions. Lead author John McBride's assertion that the predicted structures encode significant physical information has been met with skepticism, as some experts question how AI algorithms can accurately capture the complexities of protein behavior.

Moreover, the study's use of an innovative metric called effective strain to identify subtle structural changes and their impact on stability has faced scrutiny for its methodology and potential biases. The researchers' statement that significant structural changes predicted by AlphaFold2 correspond to substantial changes in stability has been met with cautious skepticism, as protein engineering necessitates rigorous validation and reproducibility.

While the research represents a significant advance in exploring the intersection of AI and protein stability, the scientific community remains cautious about the implications for protein engineering and drug development. Scientists emphasize the critical need for thorough validation, replication, and independent verification of the findings, highlighting the importance of rigor and reliability in scientific research.

As the debate regarding the IBS research on protein mutant stability continues, the scientific community remains divided on whether AI-predicted structures can genuinely enhance our understanding of protein behavior and stability. The conclusions of the study, though intriguing, have sparked skepticism and call for a comprehensive examination of the methodology, assumptions, and implications of the research within the broader context of protein engineering and biological science.

NVIDIA's quarterly Data Center revenue hit a record $26.3 billion, marking a 16% increase from Q1 and a 154% increase from a year ago

Tyler O'Neal, Staff Editor SCIENCE

The technology industry is buzzing with excitement as NVIDIA, the renowned semiconductor giant, has announced its financial results for the second quarter of fiscal 2025. NVIDIA’s figures are impressive, with a record-breaking quarterly revenue of $30.0 billion, representing a 15% increase from the previous quarter and a staggering 122% surge from a year ago.

One of the standout highlights is the record quarterly Data Center revenue of $26.3 billion, demonstrating a remarkable 16% growth from the previous quarter and an astonishing 154% surge from a year ago. NVIDIA's founder and CEO, Jensen Huang, emphasized the company's success, particularly in the data center domain, stating, "NVIDIA achieved record revenues as global data centers are in full throttle to modernize the entire computing stack with accelerated computing and generative AI."

The financial report also brings encouraging news for shareholders, with NVIDIA returning $15.4 billion in the first half of fiscal 2025 in the form of shares repurchased and cash dividends. Furthermore, the company's Board of Directors has authorized an additional $50.0 billion in share repurchases, indicating a strong commitment to rewarding its shareholders.

Looking ahead, NVIDIA forecasts a third-quarter revenue of $32.5 billion, reflecting the company's optimism for continued growth. The industry is eagerly anticipating how NVIDIA will leverage its momentum to drive further innovations and solidify its position as a leading force in the technology sector.

In addition to its financial performance, NVIDIA shared notable achievements across its diverse product portfolio, including record Gaming revenue, groundbreaking advancements in generative AI technologies, and significant developments in the automotive and robotics spheres. These accomplishments highlight NVIDIA's multifaceted approach toward technological innovation and its unwavering commitment to pushing boundaries in various domains.

The financial results have sparked curiosity among industry analysts, investors, and technology enthusiasts, prompting enthusiastic discussions on how NVIDIA's monumental achievements will shape the future of artificial intelligence, data centers, gaming, and other pivotal sectors within the technology landscape.

As the announcement captivates the attention of the global technology community, it has become evident that NVIDIA's accomplishments in the second quarter of fiscal 2025 have become a subject of eager anticipation and inquisitive admiration. Indeed, the financial results unveiled by NVIDIA are not only a testament to the company's remarkable performance but also an intriguing insight into the dynamic evolution of the technology industry as a whole.

UVA researchers claim to crack the autism code

Tyler O'Neal, Staff Editor SCIENCE

In a breakthrough that could revolutionize the diagnosis and treatment of autism, researchers from the University of Virginia claim to have decoded the neurodivergent brain. Using mathematical modeling techniques, their system can reportedly identify genetic markers of autism in brain images with remarkable accuracy. While this advancement holds potential benefits, it also raises challenging ethical concerns.

The multi-university research team, led by engineering professor Gustavo K. Rohde, has achieved accuracy rates between 89 to 95% in identifying autism-related genetic markers solely from brain scans. They argue that their approach, which focuses on genetic information rather than behavioral cues, could lead to earlier interventions and personalized medicine.

Traditional autism diagnoses heavily rely on behavioral observations, leading to delays in identification and intervention. If this new system proves effective, doctors could potentially bypass the need for behavioral assessments, allowing for quicker and more accurate diagnoses. Such progress may offer hope for individuals and families affected by autism.

However, relying solely on brain imaging and genetic markers for diagnosis and treatment raises concerns that cannot be overlooked. Critics caution that reducing autism to a solely genetic phenomenon oversimplifies the complex interplay between genetics, environment, and behavioral factors.

Moreover, the use of machine learning and mathematical modeling techniques, such as transport-based morphometry (TBM), brings its own set of challenges. While TBM claims to extract mass transport information from medical images, there are concerns about the reliability and validity of these mathematical models. Skeptics argue that this approach may overlook or misinterpret vital data that could potentially lead to misdiagnoses or inadequate treatments.

Additionally, the research team utilized data from the Simons Variation in Individuals Project, a specific group with autism-linked genetic variations. This lack of diversity in study participants raises concerns about the generalizability of the findings and the potential exclusion of individuals with different genetic profiles or diverse backgrounds.

The researchers acknowledge that understanding the biological basis of autism is a complex task but assert that their method represents a crucial step in unraveling the gene-brain-behavior relationship. However, it is crucial to approach these claims with cautious investigation and recognize the limitations of this approach.

The potential impact of this research is significant, as it aims to transform the landscape of autism diagnosis and treatment. Yet, the medical community must navigate the ethical implications of relying solely on brain imaging and genetic markers, ensuring that personalized medicine does not overlook the unique needs and experiences of neurodivergent individuals.

As we embrace the promise of this breakthrough, it is imperative that we also engage in a critical dialogue about the broader implications and potential risks associated with such advancements. Only through careful consideration and holistic approaches can we move towards a future that supports and empowers all individuals, regardless of their neurological differences.