fLPS: Fast discovery of compositional biases for the protein universe
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
BACKGROUND: Proteins often contain regions that are compositionally biased (CB), i.e., they are made from a small subset of amino-acid residue types. These CB regions can be functionally important, e.g., the prion-forming and prion-like regions that are rich in asparagine and glutamine residues. RESULTS: Here I report a new program fLPS that can rapidly annotate CB regions. It discovers both single-residue and multiple-residue biases. It works through a process of probability minimization. First, contigs are constructed for each amino-acid type out of sequence windows with a low degree of bias; second, these contigs are searched exhaustively for low-probability subsequences (LPSs); third, such LPSs are iteratively assessed for merger into possible multiple-residue biases. At each of these stages, efficiency measures are taken to avoid or delay probability calculations unless/until they are necessary. On a current desktop workstation, the fLPS algorithm can annotate the biased regions of the yeast proteome (>5700 sequences) in <1 s, and of the whole current TrEMBL database (>65 million sequences) in as little as ~1 h, which is >2 times faster than the commonly used program SEG, using default parameters. fLPS discovers both shorter CB regions (of the sort that are often termed 'low-complexity sequence'), and milder biases that may only be detectable over long tracts of sequence. CONCLUSIONS: fLPS can readily handle very large protein data sets, such as might come from metagenomics projects. It is useful in searching for proteins with similar CB regions, and for making functional inferences about CB regions for a protein of interest. The fLPS package is available from: http://biology.mcgill.ca/faculty/harrison/flps.html , or https://github.com/pmharrison/flps , or is a supplement to this article.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it