ABlooper: Fast accurate antibody CDR loop structure prediction with accuracy estimation

Antibodies are a key component of the immune system and have been extensively used as biotherapeutics. Accurate knowledge of their structure is central to understanding their function. The key area for antigen binding and the main area of structural variation in antibodies are concentrated in their six complementarity determining regions (CDRs), with the most variable being the CDR-H3 loop. The sequence and structure variability of CDR-H3 make it particularly challenging to model. Recently, deep learning methods have offered a step change in our ability to predict protein structures. In this work we present ABlooper, an end-toend equivariant deep-learning based CDR loop structure prediction tool. ABlooper predicts the structure of CDR loops with high accuracy and provides a confidence estimate for each of its predictions. On the models of the Rosetta Antibody Benchmark, ABlooper makes predictions with an average H3 RMSD of 2.45Å, which drops to 2.02Å when considering only its 76% most confident predictions.

Evaluation of Mineralogy per Geological Layers by Approximate Bayesian Computation

Summary We propose a new three-step methodology to perform an automated mineralogical inversion from wellbore logs. The approach is derived from a Bayesian linear-regression model with no prior knowledge of the mineral composition of the rock. The first step makes use of approximate Bayesian computation (ABC) for each depth sample to evaluate all the possible mineral proportions that are consistent with the measured log responses. The second step gathers these candidates for a given stratum and computes through a density-based clustering algorithm the most probable mineralogical compositions. Finally, for each stratum and for the most probable combinations, a mineralogical inversion is performed with an associated confidence estimate. The advantage of this approach is to explore all possible mineralogy hypotheses that match the wellbore data. This pipeline is tested on both synthetic and real data sets.

Confidence Predictions Affect Performance Confidence and Neural Preparation in Perceptual Decision Making.

The decisions we make are usually accompanied by a feeling of being wrong or right – a confidence estimate regarding the correctness of our decisions. The questions which information this confidence estimate is based on, and what confidence is used for, have increasingly become a focus of research into decision-making. This research has largely focused on confidence regarding current or past decisions, and successfully identified for example how characteristics of the stimulus affect confidence, and how communicating confidence can affect group decisions. Here, we report two studies which implemented a color-discrimination task which introduced a novel metacognitive measure: predictions of confidence for future perceptual decisions. Using behavioral measures, computational modeling, and EEG, we tested the hypothesis that experience-based confidence predictions are one source of information which affects how confident we are in future decision-making and that one key purpose of confidence is to prepare future encounters of a task. Results from both studies show that participants develop precise confidence predictions informed by confidence experienced in past trials. Notably, our results show a bi-directional link between predicted and experienced (performance) confidence: confidence predictions are not only informed by, but can also modulate performance confidence; this finding supports our recent proposal that confidence judgments are based on multiple sources of information, including expectations. We found further support for this bi-directional link in neural correlates of stimulus-preparation and processing. EEG measures of preparatory neural activity (contingent negative variation; CNV) and evidence accumulation (centro-parietal positivity; CPP) show that predicted confidence affects neural preparation for stimulus processing, supporting the proposal that one purpose of confidence judgments may be to learn about performance for future encounters and prepare accordingly.Taken together, our results suggest that confidence integrates information from various sources, and affects neural processing profoundly. The bi-directional link between performance confidence and predicted confidence suggests that confidence signals are exploited to increase precision in preparation and evaluation of future decisions.