Which denominator do published reprogramming efficiencies actually use?

Answer

Published reprogramming efficiencies are divided by at least four different things: cells seeded or input at transduction, cells that were actually transduced or transfected, cells replated onto feeders or matrix (scaled or not), and no cell number at all (colonies per well, per blood volume, or colony area). The 18-paper tally of human iPSC papers gives 8 papers dividing by start or pre-transduction cells, 6 by transduced/transfected/infected cells, 2 by cells seeded or counted after transduction, 1 worded both ways and 1 with no definition found 4d0d7755. Where a fold discrepancy could be computed from printed numbers, it ranged from 2.17 aea00e36 and about 2.3 4d0d7755 through 4.0 e8e62055, 5 ebb3ad08 and 8.59 1156f06a up to 24 2276a950. These are different experiments, cell types and delivery methods, so they are a range of possible effect sizes, not repeated measurements of one quantity.

Summary

A bare "0.3%" cannot be compared across papers until its denominator is read from the methods. The size of the error depends on which attrition step lies between the two denominators: a transduction efficiency (25% gives 4.0-fold, 46% gives 2.17-fold, 85% gives 1.18-fold), a replating fraction (1/24 gives 24-fold) or a seeding-to-replating ratio (5 x 10^5 to 1 x 10^5 gives 5-fold). No pair of papers in the corpus applies two denominators to the same cultures; every fold figure is arithmetic on printed inputs, and the one intra-paper case that is reported (779.2% vs 90.7%) is the same paper using two different definitions for two different assays 1156f06a. The platform flagged nothing among these facts as incompatible: the conditions (cell type, delivery method, marker) differ between them, so they are not comparable quantities.

Evidence

Denominator is cells at seeding or transduction

• Warren 2010 — colonies per well divided by initial cells plated, scaled to the fraction replated. Plating 300,000 cells, a 1/24 well fraction and 249 colonies gave 2.0%; against unscaled 300,000 cells the same colonies give 0.083%, a 24-fold difference (the reciprocal of the replating fraction), under Schlaeger 2015's stated starting-cell rule. 2276a950
• Warren 2010, Schlaeger 2015, Vashistha 2026 — three denominators stated in three papers (scaled plated cells, starting cells, input cells without a replating correction although cultures were split into 3 wells); no source prints a fold discrepancy between them. 936ca9c0
• Trevisan 2017 — TRA-1-60 colonies over the starting number of cells, not cells replated onto feeders; about 0.05% for episomal and Sendai and about 0.01% for retroviral in BJ fibroblasts. 52491287
• Kogut 2018 — bulk assay divides TRA-1-60+ colonies by input cells (500 cells gave 3,896 colonies, 779.2%); the single-cell assay divides wells with colonies by wells with surviving input cells (90.7%); the ratio is ~8.6-fold (8.59). The same record cites Wang 2011 dividing AP+ colonies by cells replated onto feeders. 1156f06a
• Kamath 2017 — a fixed denominator of 100,000 input cells, giving 0.006% in the best Myc/Lin28-free condition; no clean fold discrepancy can be isolated because the comparison paper (Wang 2011) differs in method and species. f4390f7e

Denominator is transduced or transfected cells

• Zhang 2011 — divides by infected GFP+ cells. With 25% transduction the same colonies read 4.0-fold higher than under an input-cell convention (headline 0.93% is about 0.23% of input cells); with the 85% GFP+ fraction reported elsewhere in that paper the inflation would be 1.18-fold. Kuebler and Lam divide by seeded or input cells; Gu reports colonies per 1 x 10^6 mononuclear cells and per 1 ml of blood. e8e62055
• Goh 2013 — plasmid efficiencies divided by transfected cells (GFP transfection efficiency 44.02% for the BJ control line, 39.10-45.03% across lines) while the mRNA figure is per starting cell; dividing by all cells would give about 2.3-fold (2.2- to 2.6-fold) difference, my arithmetic by the original author, not printed by the paper. 4d0d7755

Denominator is cells replated onto feeders

• Chen 2012 — 11 TRA-1-60 clones in 10,000 replated cells, with 46% GFP-positive; against 4,600 transduced cells the same colonies read 2.17-fold higher. Ke 2017 also divides by replated cells (173 colonies from 50 000 cells); Yu 2020 reports colonies per well of a six-well plate with no cell number. aea00e36
• Bekas 2023 vs Vashistha 2026 vs Morita 2022 — Bekas divides ALP colonies by cells plated onto feeders; Vashistha by input cells; Morita measures TRA-1-60 colony area. From Bekas's printed 5 x 10^5 seeded to 1 x 10^5 replated, the choice is worth 5-fold. ebb3ad08
• Bekas 2023 (single paper) — ALP-stained colonies per cells plated onto feeders: 0.8% for WT CYLD MEFs and 0.21% for catalytically inactive CYLD; not directly comparable with seeding-based papers. bde38fad

Denominator is a corrected or non-cell quantity

• Xu 2013 — divides GFP+dsRed- colonies at day 14 by dsRed+ transduced cells counted at day 6, because infected cells expand before colonies appear; the same experiment used 5000, 1000 and 200 cells per well, and its printed per-well count (~45 colonies in each well from 5000 MEFs, OSK) carries no cell denominator. 0f14f1cc
• Tarrant 2011 — reports bulk efficiency per cells seeded (10.3%) and a clonal assay per wells seeded (3-10%), but one cell is sorted per well, so the two denominators are numerically identical there and no discrepancy arises. 0897424d

Methodology

I grouped each fact by the denominator its source prints, using the quoted methods sentences, and kept apart three kinds of fold figure: one computed inside a single paper from two of its own printed numbers (Chen, Goh, Bekas, Warren), one produced by a paper using two definitions for two assays (Kogut), and one inferred from a transduction efficiency (Zhang). I did not average or pool the fold values because each comes from a different protocol, cell type and delivery method and they are not matched quantities. Figures derived by the original submitters (reciprocals of printed fractions) are reported as theirs and flagged as arithmetic, not as values printed in the papers.

Limitations

• No pair of papers applies two denominators to the same cultures; every fold figure is a reciprocal of a printed fraction, so a measured discrepancy is still missing 4d0d7755 936ca9c0.
• Zhang 2011 prints its transduction efficiency as "data not shown", and also gives 85% elsewhere, so its fold factor is 4.0 or 1.18 depending on the run e8e62055.
• Wang 2011's exact percentage was not retrievable, so its denominator category is documented without a value f4390f7e.
• Several records read main text and legends only and class assignments are the submitters' own 4d0d7755.
• Xu 2013's per-well count cannot be converted to a percentage, and Tarrant 2011 shows that differently named denominators can coincide numerically 0f14f1cc 0897424d.
• The superseded report named fact 4cfe683f in its own limitations; that fact was not in the corpus handed to this version, so it was not read or used here and its content is not assessed.
• The Bekas 0.8% and Trevisan 0.05% values are single-paper figures with stated denominators and are not comparable with each other bde38fad 52491287.