Investigates shapeshifting proteins with computation and experiment. Tenure-track. Mother of two. Foodie. Yoga lover. Opinions are mine.
Jan 8, 2024 • 14 tweets • 5 min read
Random sequence sampling outperforms sequence clustering for the fold-switching proteins presented in the AF-cluster Paper (). Observations of data directly from the AF-cluster Repo led us to this realization. 🧵nature.com/articles/s4158… biorxiv.org/content/10.110…
Observation 1: The Paper claims that AF2 generates higher confidence models of RfaH's autoinhibited state from sequence clusters than from the full multiple sequence alignment (MSA). No. Autoinhibited models directly from the Repo have lower confidences than full-MSA preds.
Nov 27, 2023 • 13 tweets • 4 min read
claims that ColabFold uses coevolutionary inference to predict alternative conformations of proteins from shallow sequence clusters. These clusters are unnecessary: ColabFold predicts the same conformations from single sequences.🧵 nature.com/articles/s4158… biorxiv.org/content/10.110…
Most predictions focused on homologs of KaiB. On average, single-sequence predictions of KaiB homologs were within 2.0 Å root-mean-squared deviation (RMSD) of their respective AF-cluster predictions. Single-sequence predictions of RfaH and Mad2 were also similar to AF-cluster.
