gladstone-institutes/clustOpt: v1.3.0

Read full text Summarise with VIRAT
Authors: Natalie Gill, Ayushi Agrawal, reubenthomas
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

New features clust_opt() gains a checkpoint_dir argument for checkpoint/resume of long runs. When set, each holdout subject's result is written to disk as it completes and reloaded on a rerun pointed at the same directory, so a run killed by an HPC wall-clock timeout resumes instead of restarting from zero. The sketched input and run seed are persisted too, so a resumed run is bit-identical to an uninterrupted one. All checkpoint I/O happens in the main process, never in future workers, so there is no parallel file-connection contention. Reusing a directory with a different configuration, input, or clustOpt version is an error, so checkpoints are never mixed across algorithm versions. Enabling checkpointing switches to deterministic per-subject seeding, so results differ from a non-checkpointed run but are stable across resumes. Default checkpoint_dir = NULL leaves existing behavior unchanged. clustOpt 1.2.4 Bug fixes ranger's "Growing trees.." and "Predicting.." progress lines are no longer printed regardless of verbose. They are now gated behind a new verbosity level 4, which covers output from packages other than Seurat (level 3 remains Seurat-only). Long runs at the default verbose = 0 are now silent.

Fields of Study & Subjects

Single-cell and spatial transcriptomics Advanced Clustering Algorithms Research Scientific Computing and Data Management

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
26 March 2026
Language
EN