Automated identification of cell-type–specific genes and alternative promoters
Bibliographic record
Abstract
Abstract Background Identifying key transcriptional features, such as genes or transcripts, involved in cellular differentiation remains a challenging problem. Current methods for identifying key transcriptional features predominantly rely on pairwise comparisons among different cell types. These methods also identify long lists of differentially expressed transcriptional features. Combining the results from many such pairwise comparisons to find the transcriptional features specific only to one cell type is not straightforward. Thus, one must have a principled method for amalgamating pairwise cell type comparisons that makes full use of prior knowledge about the developmental relationships between cell types. Method We developed Cell Lineage Analysis (CLA), a computational method which identifies transcriptional features with expression patterns that discriminate cell types, incorporating Cell Ontology knowledge on the relationship between different cell types. CLA uses random forest classification with a stratified bootstrap to increase the accuracy of binary classifiers when each cell type have a different number of samples. Regularized random forest results in a classifier that selects few but important transcriptional features. For each cell type pair, CLA runs multiple instances of regularized random forest and reports the transcriptional features consistently selected. CLA not only discriminates individual cell types but can also discriminate lineages of cell types related in the developmental hierarchy. Results We applied CLA to Functional Annotation of the Mammalian Genome 5 (FANTOM5) data and identified discriminative transcription factor and long non-coding RNA(lncRNA) genes for 71 human cell types.With capped analysis of gene expression (CAGE) data, CLA identified individual cell-type–specific alternative promoters for cell surface markers. Compared to random forest with a standard bootstrap approach, CLA’s stratified bootstrap approach improved the accuracy of gene expression classification models for more than 95% of 2060 cell type pairs examined. Applied on 10X Genomics single-cell RNA-seq data for CD14 + monocytes and FCGR3A + monocytes, CLA selected only 13 discriminative genes. These genes included the top 9 out of 370 significantly differentially expressed genes obtained from conventional differential expression analysis methods. Discussion Our CLA method combines tools to simplify the interpretation of transcriptome datasets from many cell types. It automates the identification of the most differentially expressed genes for each cell type pairs CLA’s lineage score allows easy identification of the best transcriptional markers for each cell type and lineage in both bulk and single-cell transcriptomic data. Availability CLA is available at https://cla.hoffmanlab.org . We deposited the version of the CLA source with which we ran our experiments at https://doi.org/10.5281/zenodo.3630670 . We deposited other analysis code and results at https://doi.org/10.5281/zenodo.5735636 .
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 | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".