More Like this

1. Mechanistic Mode Connectivity

   Lubana, Ekdeep Singh; Eric J. Bigelow; Robert P. Dick; David Krueger; Hidenori Tanaka. (July 2023, Proc. Int. Conf. on Machine Learning)

   We study neural network loss landscapes through the lens of mode connectivity, the observation that minimizers of neural networks retrieved via training on a dataset are connected via simple paths of low loss. Specifically, we ask the following question: are minimizers that rely on different mechanisms for making their predictions connected via simple paths of low loss? We provide a definition of mechanistic similarity as shared invariances to input transformations and demonstrate that lack of linear connectivity between two models implies they use dissimilar mechanisms for making their predictions. Relevant to practice, this result helps us demonstrate that naive fine-tuning on a downstream dataset can fail to alter a model’s mechanisms, e.g., fine-tuning can fail to eliminate a model’s reliance on spurious attributes. Our analysis also motivates a method for targeted alteration of a model’s mechanisms, named connectivity-based fine-tuning (CBFT), which we analyze using several synthetic datasets for the task of reducing a model’s reliance on spurious attributes. 

   more » « less
2. EDA photochemistry: Mechanistic investigations and future opportunities

   https://doi.org/10.1016/j.chempr.2023.06.013  

   Wortman, Alan K.; Stephenson, Corey R.J. (September 2023, Chem)
3. Stereodivergent Zweifel Olefination and its Mechanistic Dichotomy

   https://doi.org/10.1002/anie.202309136  

   Manna, Samir; Paul, Swagata; Kong, Wang‐Yeuk; Aich, Debasis; Sahoo, Rupam; Tantillo, Dean J.; Panda, Santanu (September 2023, Angewandte Chemie International Edition)

   Abstract Stereoselective Zweifel olefination using boronate complexes carrying two different reactive π‐systems was achieved to synthesize vinyl heteroarenes and conjugated 1,3‐dienes in good yield and up to 100 % stereoselectivity, which remains unexplored until now. Most importantly, we report the unprecedented formation ofEvs.Z‐vinyl heteroarenes for different heteroarenes under identical conditions. Density functional theory (DFT) investigations unveil the mechanistic dichotomy between olefin and heteroarene activation followed by 1,2‐migration, leading toEorZ‐vinyl heteroarenes respectively. We also report a previously unknown reversal of stereoselectivity by using 2,3‐Dichloro‐5,6‐dicyano‐1,4‐benzoquinone (DDQ) as an electrophile. The Zweifel olefination using a boronate complex that carries two different olefins was previously unexplored due to significant challenges associated with the site‐selective activation of olefins. We have solved this problem and reported the site‐selective activation of olefins for the stereoselective synthesis of 1,3‐dienes. 

   more » « less
4. Iron Olefin Metathesis: Unlocking Reactivity and Mechanistic Insights

   https://doi.org/10.1021/jacs.4c04356  

   Lincoln, Zachary S; Iluc, Vlad M (June 2024, Journal of the American Chemical Society)

   Olefin metathesis catalyzed by iron complexes has garnered substantial interest due to iron’s abundance and nontoxicity relative to ruthenium, yet its full potential remains untapped, largely because of the propensity of iron carbenes to undergo cyclopropanation instead of cycloreversion from a metallacycle intermediate. In this report, we elucidate the reactions of [{PC(sp2)P}Fe(L)(N2)], ([PC(sp2)P] = bis[2-(diisopropylphosphino)phenyl]methylene) with strained olefins, unveiling their capability to yield metathesis-related products. Our investigations led to the isolation of a structurally characterized metallacyclobutane during the reaction with norbornadiene derivatives, ultimately leading to a ring-opened iron alkylidene. These findings provide compelling evidence that iron complexes adhere to the Chauvin olefin metathesis mechanism. 

   more » « less
5. Characterizing and Predicting Protein Hinges for Mechanistic Insight

   https://doi.org/10.1016/j.jmb.2019.11.018  

   Khade, Pranav M.; Kumar, Ambuj; Jernigan, Robert L. (January 2020, Journal of Molecular Biology)