Bio+MedVis Challengebookmark_border

Adaptive molecular dynamics (MD) ensembles capture complex conformational behavior, but interpreting Markov State Model (MSM) assignments across many trajectories, frames, states, and ligand conditions remains challenging. We present ConformaLens, a linked visual analytics application for comparing free, TMP-bound, and SPA-bound Aβ42 simulations. Its coordinated timeline, transition footprints, interaction glyphs, and 3D structural views connect ensemble-level MSM dynamics with ligand effects and residue-level evidence, revealing representative states and residue-residue or residue-ligand differences across trajectories and frames.

While adaptive molecular dynamics successfully escapes local minima to capture rare conformational states, it generates a branching diversity of trajectories, rendering aggregate metrics insufficient for comprehensive analysis. CôôrdVis-MD, a web-based visual analytics environment, was designed to tackle this issue. Four coordinated views connect trajectory evolution, features, events, conformational states, and molecular structure via real-time cross-filters. Cohort beeswarms and lineage trees reveal simulation hierarchies, enabling the selection of ancestors, descendants, and transition cohorts over time. 3D spatial conformations are explored through side-by-side timelines, panoramic small multiples, and hierarchical treemaps. Finally, we inspect structural features globally or individually via compact heatmaps. As a result, we present two use cases that reveal adaptive simulation artifacts within the CoVAMPnet data.

Adaptive molecular dynamics generates thousands of short, unequal trajectories labeled by abstract Markov state identifiers, yet the scientific question is whether a drug candidate alters the conformational ensemble of a disordered peptide. We present Ensemble Lens, an interactive comparative visualization in which every view places the three simulated systems side by side on shared scales, enabling readers to interpret abstract state labels as shapes, interactions, and mechanisms. Two principles guide the design: encoding measurements relative to a reference system and subtracting an explicit chance-contact baseline before attributing any interaction to binding. Applied to the CoVAMPnet Aβ42 dataset, the visualization reveals differences between tramiprosate and its metabolite SPA that are obscured by state populations alone.

MammoWeave is a provenance-preserving visual-analytics workbench for auditing mammography datasets and automated labels before model development. We introduce three coupled design contributions: an evidence-role grammar that keeps source pixels, acquisition metadata, reference labels, automated scores/boxes, and derived audit measurements visually and semantically distinct; a fidelity-first cohort-to-pixel workflow that connects population context, interpretable quality structure, threshold-conditioned score-label discordance, and paired multi-resolution evidence; and a Casebook handoff that retains source, transformation, uncertainty, and review status for selected findings. These ideas are operationalized as a five-stage process---Survey, Stratify, Contradict, Inspect, and Record---with coordinated selection, alternative encodings, a review-threshold control, and paired-image inspection. A walkthrough on the supplied Challenge 2 archives shows how aggregate signals can be traced back to source products while exposing metadata drift, sparse standard-view context, qualified score-label discordance, and cross-resolution coordinate incongruence. The prototype intentionally bounds interpretation and interaction, but the visual contribution is the persistent evidence structure that prevents derived findings from silently becoming clinical or annotation verdicts.

Adaptive molecular dynamics produces many short simulations that begin from frames of earlier runs. This branching improves sampling, but it makes a simple question hard to answer: when a state-transition probability changes, which simulations produced the change? StateWeaver joins state occupancy and transition comparisons with the simulation lineage. It covers the supplied assignments for free Aβ42, Aβ42 with tramiprosate, and Aβ42 with 3-sulfopropanoic acid: 955 simulations and 1,899,875 assigned frames in each of three Markov-state models. Selecting a transition shows the treatment and baseline probabilities, their difference, a descriptive standard error, event counts, and the runs in which the transition occurs. In the 3-state model, selecting SPA 0→1 reduces 319 simulations to 16 runs containing 209 observed transitions. The difference is +0.10 percentage points with a standard error of 0.18, so it remains uncertain. StateWeaver provides a focused way to move from an ensemble-level difference back to the simulations behind it without turning a descriptive pattern into a biological claim.