21 of 30 contrast × method slots have a scored model (held out: replicate rep3); best held-out AUC 0.926 (PD-L1 protein high vs low · new screen 1 all k-mer); best regression Spearman ρ 0.743 (PD-L1 protein level (regression) · combined (hash + new screen 1, all k-mer)); the combined hash + new screen 1 model beats the hash model in 5 of 5 contrasts

21 models over 5 contrasts, scored on held-out replicate rep3

Dataset: SRP268462 / GEO GSE153056 · CITE-seq (ECCITE-seq) of THP-1 cells, one acute monocytic leukemia cell line, not patient tumour · curated per-cell guide labels and per-cell surface-protein (ADT) counts · 8 SRA runs · 20,729 cells in the metadata · 3 replicates. Counted over: every contrast × method with an output under the S3 prefixes sent for this page. Each contrast keeps only the cells of its classes. The model is trained on replicates rep1 and rep2 and scored on rep3.

Each row is one model. hash is the P2G stage 3 model; new screen 1 ranks targets by presence ratio between the classes; combined fits one lasso on the union of both candidate sets; combined dict averages an all-k-mer model and a SPLASH model. Binary contrasts report held-out AUC; regressions report Spearman ρ (rank agreement of predicted and measured values), Pearson r (straight-line agreement, the measure most RNA→protein papers use) and R² (share of the variance explained compared with predicting the mean); the table under the runs lists all three.

Regression models: Spearman ρ, Pearson r and R² on held-out rep3 cells (5,516)

Pearson r is shown beside Spearman ρ because most RNA→protein papers report Pearson. Both come from the engine’s own metric files on the same held-out cells; the hash all k-mer run saved only Spearman ρ and R², so its Pearson (†) was recomputed afterwards from the final model’s saved coefficients on the same test cells; that rebuild reproduces the engine’s own Spearman to 4–6 decimals. R² below 0 means the model predicts worse than the training average.

Papalexi et al. 2021 reports no classifier result, so no published value sits beside P2G's

PD-L1 protein level from RNA: R², Spearman ρ and Pearson r on held-out rep3 cells (5,516)

The authors give no R² or correlation for PD-L1 protein vs RNA, so P2G's R² 0.55 has no published counterpart. The nearest reference points are our own RNA baselines from the study index. A gene-count elastic net reaches ρ 0.778, a little above P2G's 0.743. CD274 mRNA alone reaches only ρ 0.396, so most of the PD-L1 protein signal comes from the IFN-γ-response state rather than from the PD-L1 transcript. That fits the authors' finding of post-transcriptional regulation. Checked against the full text (PMC8011839) on 2026-10-09.

Two kinds of label, neither a cell type: CRISPR guide identity (the KO contrasts) and one protein measurement, PD-L1 (the PD-L1 contrasts); every cell is a THP-1 monocytic cell

Counted over: the cells of each contrast, joined to the exact train and test cell IDs in each run's run_split.rds. A "KO" cell carries a guide against that gene whether or not the knockout worked: the authors' mixscape KO calls are not deposited, so escaping cells stay in the KO class. The PD-L1 contrasts use the PD-L1 antibody tag of each cell: the state contrast compares the top third with the bottom third, and the level contrast predicts the value itself.

Stage 1, all anchors: 17 to 212 of about 64,000 anchors pass |r| > 0.2

Counted over: every anchor in each hash run's stage1_pearson_all.tsv.gz, before any selection. Passes per contrast: IFN-γ pathway KO vs NT 17 · STAT1 KO vs NT 37 · IRF1 KO vs NT 18 · PD-L1 protein high vs low 212 · PD-L1 protein level (regression) 196. r is computed on read count × f, so its sign follows f; the page shows |r|. An anchor's gene is the gene of its top target in stage2_cv_pass.tsv.gz.

New screen 1 keeps targets present in enough cells, ranks them by presence ratio, and picks the cut K by training cross-validation

Counted over: every anchor-target pair in each contrast's new screen 1 run, read from its run.log. Exclusive targets are present in one class only; graded targets are present in both and ranked.

Selected sequences: 10,646 of 17,399 selected rows (summed over the 21 models) match the genome exactly

Counted over: the sequences each model selected (non-zero weight): 17,399 across the 21 models, out of 24,079 rows in their tables. 3,235 of them have no target recorded.

Selected genes vs expression: how much of each model's weight sits on highly expressed genes

Counted over: the genes the selected sequences are drawn on, per model, weighted by summed |weight|. Expression: GSM4633614_ECCITE_cDNA_counts.tsv.gz (GEO GSE153056), normalised to counts per 10,000 per cell.

Literature: what PubMed says about each heavy gene with IFN-γ signalling and PD-L1 / checkpoint

Counted over: the 15 heaviest named genes of each model. PubMed title/abstract searches: gene + IFN-γ signalling, gene + PD-L1 / checkpoint, and gene + both (a paper that names the gene and both topics).

120 alternative-splicing, 904 mismatch (1 or more) and 516 BLAST-only features (5 of them imperfect) among the selected sequences of the 21 models

Counted over: the selected sequences (non-zero weight) of every model, after the 1% rule: a sequence present in fewer than 1% of training cells is left out (98 selected sequences, almost all from hash models; new screen 1 already requires at least 5%). (a) crosses a STAR junction that is not in the reference transcript; compatible transcripts have that exact intron, incompatible ones span it but splice differently or keep it. (a) also counts a soft-clipped sequence that is full length only on alternative isoforms. (b) edit distance of 1 or more from the aligner that placed it, on a full-length placement (split into 1 and 2 or more); soft-clipped and unplaced sequences have no comparable edit distance and are left out of (b). (a) or (b) is counted once per sequence, with the overlap shown, over the selected sequences and over all stage 3 rows. This follows section 3h of the GSE123813 instructions (Julia, 2026-10-08). (c) placed by neither aligner, but hit by local BLAST (GENCODE, genome, IPD-IMGT/HLA) or NCBI core_nt; perfect = 100% coverage and 100% identity. A sequence can be in more than one group. The same features for every study are collected on the running feature ledger.

Each selected sequence drawn on its gene, marked by how it matches the genome and how other isoforms read it

Drawn: the selected sequences with a genome position. Listed in the table under the track: every row on that gene, selected or not.

Choose a model, then a gene. Genes are listed by the total weight of their selected sequences. Click a mark or a table row for its card.

2,282 selected sequences (summed over the 21 models) did not align confidently and were BLASTed locally

Counted over: selected sequences of the chosen model (section 4's model menu). The pool is every sequence without a confident full-length alignment with at most 1 mismatch; sequences with no target recorded are counted but cannot be searched.

Every sequence with 2+ mismatches or soft-clipping, BLASTed against mRNAs, the whole genome and HLA alleles

Counted over: every row with 2 or more mismatches (full length) or any soft-clipping, selected or not. Coverage = aligned bases ÷ sequence length; imperfect values are amber.

5c · Selected sequences never BLASTed, and the ones at a known SNP or with a base change

Counted over: the selected sequences of the model chosen in section 4.

Selected sequences at a known dbSNP site, a ClinVar or COSMIC record, or with a base change called by P2G

Counted over: the selected sequences of the model chosen in section 4 (heaviest 60). Known variants in HLA and Fc-gamma genes, for every model, are in section 6c.

Do the selected sequences fall on the knocked-out genes, on PD-L1, or on interferon-response genes?

Counted over: the selected sequences of every model, by summed |weight|. Groups: the contrast's knocked-out genes, CD274 and PDCD1LG2, MSigDB HALLMARK_INTERFERON_GAMMA_RESPONSE (M5913) and HALLMARK_INTERFERON_ALPHA_RESPONSE (M5911). The lookup counts every row, selected or not. The guide check compares every row of every model with the authors' guide spacers (49 matches in all). HLA and Fc-gamma genes, which differ between people, are counted per model in section 6c.

Immune genes that vary between people: how much of each model’s score is in HLA and Fc-gamma genes?

Why this matters: a model can learn which person a cell came from instead of which condition it is in, and HLA is the most variable region of the human genome. All cells here come from one cell line (THP-1), that is from one person, so weight on these genes cannot come from differences between people. HLA genes are switched on by interferon gamma, the pathway these knockouts disrupt, so here they can be real signal.

3 of 21 models are flagged: HLA and Fc-gamma genes make up 5% or more of their score.

4 models select no sequence in these genes: STAT1 KO vs NT · new screen 1 SPLASH; IRF1 KO vs NT · hash all k-mer; PD-L1 protein high vs low · new screen 1 SPLASH; PD-L1 protein level (regression) · new screen 1 SPLASH.

The selected sequences in these genes, model by model

Click a model to open its sequences. 17 models have at least one.

Checks run on this build (they inform, they never block)

Counted over: every model table read for this page.

Annotation coverage per model (rows with a value)

Every file this page was built from, with its S3 path

Read: the files below, the GEO expression matrix and guide barcodes, GENCODE v46, GRCh38.p13, IPD-IMGT/HLA 3.65.0, NCBI gene_info and MSigDB hallmark sets.

Every rule this page followed

Source

Built to match the GSE123813 sequence-review page (artifact 6ChpFqMvKXfBMEbrJT1LsV) and the page instructions in rosabiocom/rosa-skills (population-comparison-site, *_PAGE_INSTRUCTIONS.md). The original build scripts are not in the skill, so this page was rebuilt: Julia's page code (renderer, gene_track_annot.js, styles) is reused, and every data file is regenerated by the shared build in website-artifact/sites/study-build (build_rows.py, run_blast.sh, blast_classify.py, expr_stats.py, lit_fetch.py, build_page.py, assemble.py).

Models

One page per study, every contrast on it. Each contrast has up to six models, chosen in every dropdown: hash all k-mer and hash SPLASH (P2G stage 3, stage3_annotated_target.tsv.gz of each dictionary), new screen 1 all k-mer and new screen 1 SPLASH (presence-ratio final-selected tables), combined (Combined_selected_seq: hash + new screen 1 candidates in one lasso) and combined dict (all-k-mer + SPLASH dictionaries, equal-weight). A slot without an output is listed in section 1 with the reason.

Scores

Binary contrasts: held-out AUC (hash: roc_auc of test_evaluation.tsv at feature_level = anchor_target; the anchor-level AUC of summary_row.tsv is shown for reference). Regressions: Spearman ρ between predicted and measured values on held-out cells, and R², the share of the variance explained compared with predicting the mean.

Weight and direction

hash: lasso_coef × f; the other three: the model coefficient. Positive pushes toward the first class of the contrast, negative toward the second. For a regression, positive raises the predicted value and negative lowers it.

Rows

One aligner per sequence (minimap2 when it placed the sequence, otherwise bowtie2); CIGAR, edit distance and cs all from that aligner. A dotted or empty cell is "none". Gene = GENCODE v46 gene overlapping the drawn position, protein-coding before lncRNA before pseudogene, named before ENSG-only. Rows with no target sequence are "no target recorded" and never placed or searched. SPLASH features of the combined-dictionary model take their sequence from the SPLASH new screen 1 table when it has one.

Isoforms

GENCODE v46 comprehensive. The reference row is MANE Select, else Ensembl canonical. An isoform row is kept only when it reads a selected sequence differently (stopping short is not a different reading); identical readings are collapsed; at most 8 rows; every transcript is accounted for in the caption.

Junctions

STAR star_junction matched exactly to the introns of every GENCODE v46 transcript: reference, annotated alternative, or not annotated (new pairing, one new site, no annotated site).

BLAST

Local first: GENCODE v46 transcripts (blastn-short, dust off, E ≤ 1, ≥ 90% identity, up to 500 targets), GRCh38.p13 top level with alternate haplotypes and patches (word size 11, E ≤ 10, ≥ 95% identity) and IPD-IMGT/HLA 3.65.0. Pool = no confident full-length alignment with ≤ 1 mismatch. Rescue bar: ≥ 30 bases at ≥ 97% identity; transcripts first, then the genome (E ≤ 1e-4). A tie across genes is ambiguous; a genome locus is a position, never a gene. Every sequence no local search explains, and every 5b "no closer match" or "clipped bases unexplained" sequence, was then searched remotely in NCBI core_nt (megablast, word size 16, E ≤ 1e-3, up to 25 targets); the card shows the best hit, its organism and the tie count. Other pool sequences were not sent remotely; a sequence still unsearched would be marked "would need a remote search".

Expression

GEO GSM4633614_ECCITE_cDNA_counts.tsv.gz (GEO GSE153056), CP10K per cell. Each model's two groups are its own train + test cells from run_split.rds (a regression is split at the median of the response). Matrix columns are matched to P2G cells by barcode. Ranks are among genes expressed in the two groups together; a GENCODE name missing from the matrix is matched through NCBI synonyms.

Literature

PubMed title/abstract counts for gene + IFN-γ signalling, gene + PD-L1 / checkpoint, and both, for the 15 heaviest named genes of each model (symbol, full name, NCBI synonyms of 4+ characters with a digit or hyphen). Up to three papers per gene with the abstract sentence that names the gene, verbatim. Each linked paper has a one-line conclusion written from its title and abstract only, with a relevance tag (central, finding, gene list only, unrelated match, no abstract).

Gene census

Groups: the contrast's knocked-out genes, CD274 and PDCD1LG2, MSigDB HALLMARK_INTERFERON_GAMMA_RESPONSE (M5913) and HALLMARK_INTERFERON_ALPHA_RESPONSE (M5911). The lookup counts every row, selected or not.

Literature: splicing, mismatches and BLAST per gene

Requested 2026-10-08. For each of the 15 heaviest genes in the literature table (3f), two columns summarise that gene's selected sequences (non-zero weight) in the chosen model: alternative splicing (STAR junctions classed against GENCODE v46, plus sequences exonic only in non-reference isoforms and the mRNA-search junction verdict) and mismatches · BLAST (edit-distance strata 0 / 1 / 2+, soft-clipped and unplaced counts, the local BLAST verdicts, and NCBI core_nt hits). A sequence belongs to the gene when it lies in the gene's track, or its stage-3 gene_name, junction annotation or BLAST transcript hit names the gene.

Feature report (section 3g) and the running ledger

Requested 2026-10-08. Every rebuild runs feature_report.py on the selected sequences (non-zero weight) of every model and keeps the ones in three groups: (a) alternative splicing, a STAR junction not in the reference transcript, with the GENCODE v46 transcripts compatible with it (that exact intron) and incompatible (span it, splice differently or keep it); (b) 1 or more mismatches from the aligner that placed the sequence; (c) no aligner placement but a BLAST hit (local, then NCBI core_nt), perfect or imperfect. The 1% rule is applied: a sequence present in fewer than 1% of training cells is left out (hash rows: Train_prev weighted by class size; new screen 1 already requires 5%; combined rows take the share of the same sequence in the hash or new screen 1 table). build_ledger.py then collates every study into the running feature ledger with study, phenotype, label basis, model, gene, direction, target rank, coefficient, weight and share of cells.

Not done

Cell-type marker status is not computed: both classes are THP-1 cells, so it cannot separate them. LitVar2 variant papers were not run. Never downloaded raw reads.

Published study

Papalexi et al. 2021 (PMID 33649593, PMC8011839) reports no AUC, sensitivity or specificity for guide identity or PD-L1 state.