GPU-Accelerated Protein Family Identification for Metagenomics
Bibliographic record
Abstract
The clustering of putative protein/Open Reading Frame (ORF) sequences available from large-scale metagenomics survey projects is a core analytical function that has led to the identification and characterization of novel protein families of environmental microbial communities. The implementation of this function, however, is currently challenged not only by data size but also by data complexity. In this paper, we present a CPU-GPU implementation of a randomized graph clustering heuristic called Shingling, which was originally developed by Gibson et al. Our implementation uses the CPU and GPU for different stages of computation, using GPUs for the most time-consuming steps. Experimental results of a 2M ocean metagenomics data set obtained from the Sorcerer II Global Ocean Sampling project show that our new implementation is able to achieve a ~7X speedup over our serial implementation without using asynchronous CPU-GPU communication, with the GPU part alone contributing to over ~374X speedup in the accelerated part. Qualitative evaluation of the 2M data set shows that our method is able to improve sensitivity of clustering over existing methods, and is more successful in recruiting more sequences into the clustering without impacting the overall specificity. As a demonstration of a large scale run, we were able to cluster a real world homology graph, containing 11M vertices and 640M edges, and constructed from sequences of an ongoing Pacific Ocean metagenomics survey project, in about 94 minutes.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".