exercise · intermediate · partly written
Model a single bacterial cell, end to end
Build the smallest living cell on paper, count its parts, its unknown genes, its interaction space and its physics, and find where the model stops being knowledge.
What this exercise asks
You build the model of the smallest living cell on paper, step by step. Each step ends in a measured answer or in a place where there is none, and the point is where the model stops being knowledge. It asks two of his questions: could a single bacterial cell be modelled end to end, and what would the molecules do if you ignored survival and only let them move. 12
Reported, not yet replicated
What end to end would mean, and what exists more
The 2022 syn3A model has 493 genes, 452 of them protein-coding, over 77,000 proteins, 7,765 molecule types and over 7,200 reactions. It is spatial for 20 minutes of cell time and well-stirred for a full cycle. The first whole-cell model (Mycoplasma genitalium, 525 genes) was called a first draft by its authors. The E. coli model has 1,214 genes. A 2026 paper reports a full cycle for syn3A; it was known to us only from the abstract and press, so we say nothing about what it omits. 134
Reported, not yet replicated
The unknown genes: his figure checked more
In syn3A, 91 of 452 protein genes were of unclear function in 2019 (30 essential, 32 quasi-essential, 29 non-essential), and the model's authors restate 87 in 2022. His 70 to 90 matches that count under that scheme, but no source defines a set of that size. In syn3.0, 149 of 473 genes were generic or unknown, which is a different scheme. In E. coli, 1,600 of 4,623 (34.6%) lacked experimental evidence in 2019 and 738 of 4,747 (15.5%) were 'uncharacterized' in 2024. 'Unknown' means three different things: no activity at all, a recognisable type without a substrate (generic), and essential but unexplained (the 30). 21567
Reported, not yet replicated
Deep problem or annotation backlog? more
Toward a missing mechanism: the essential unclear genes are essential, the first minimal design failed because single-gene lists missed redundancy, and four of the seven genes added for division are membrane-associated proteins of unknown function (abstract only). Toward a backlog: the unknown class leans to membrane proteins that are hard to assay (49% with predicted helices versus 6% of well-annotated types), and computational methods already assign a function to about half. The evidence does not settle it, because the resolutions so far are predictions, not bench results. His reading is on the linked question page in his words; this page does not correct it, it adds the sourced account beside it. 82910
Contested
The interaction space against what has been tested more
For n protein genes the k-gene combinations number C(n,k). For 452 genes: 101,926 pairs, 15,288,900 triples, 1,716,179,025 quadruples. Against that, one yeast screen tested about 23 million pairs and another 195,666 triples; the E. coli screens tested about 600,000 and 315,400 pairs. Measured density of interacting pairs runs from 0.6% to 7.1% in E. coli alone. None of this exists for syn3A. 111121314
Our inference
The physics-only view more
If molecules only diffuse, a protein crosses a 0.4 um cell in milliseconds (our arithmetic) and meets a single partner copy in a fraction of a second to a few minutes, depending on how reactive the binding patch is. A purely physical account then fails in three places: it mispredicts how fast proteins move in cytoplasm without a fitted term, it cannot say which molecules react, and for syn3A the key physical input is borrowed. Diffusion is computed; reactivity, stickiness and the cell's energy state are assumed or measured elsewhere. 1516171
Our inference
What this page did and did not read deep
Opened in full or in large part: the 2022 syn3A model, the 2019 essential-metabolism paper and its supplement, the 2024 E. coli y-ome paper (through a summarising tool), the diffusion review and the two crowding papers. Abstract or supplement-page only: the 2021 division paper, the yeast and E. coli screen pages for the claims quoted. Not read: the 2026 whole-cell paper. Figures marked inferred are our own arithmetic from the inputs given beside them. 12715912
Reported, not yet replicated
The task
Model a single minimal bacterial cell end to end, and find exactly where the model stops being knowledge.
What you are given
The cell
JCVI-syn3A: 543,000 DNA base pairs, 493 genes, of which 452 make proteins. It is a laboratory-made minimal bacterium descended from a Mycoplasma genome. For scale, the first free-living genome of this kind, Mycoplasma genitalium, had 580,070 base pairs and about 470 coding regions, 152 of them (32%) with no assigned role in 1995. 118
Reported, not yet replicated
What the 2022 model contains
Over 77,000 proteins, 503 ribosomes placed from cryo-electron tomograms, 7,765 molecule types and over 7,200 reactions. It transcribes all 493 genes and translates all 452 messenger RNAs. The spatial version covers 20 minutes of cell time, without growth or DNA replication, at 8 to 10 GPU-hours per 20 simulated minutes. 1
Reported, not yet replicated
The measured doubling time
About 105 minutes. The well-stirred version of the model gives 97 to 98 minutes. 1
Reported, not yet replicated
Unknown-gene counts, each under its own rule
JCVI-syn3.0 (2016): 149 of 473 genes (31%), made of the classes generic (84) and unknown (65). JCVI-syn3A (2019): 91 of 452 protein genes of unclear function, of which 30 are essential, 32 quasi-essential and 29 non-essential. JCVI-syn3A (2022, restated by the model's authors): 87 of 452 (20%). E. coli K-12 (2019): 1,600 of 4,623 genes (34.6%) with no experimental evidence of function. E. coli K-12 (2024, EcoCyc): 738 of 4,747 (15.5%) uncharacterized, plus 1,247 in a new partial tier. These are not one quantity measured twice; each is a count under a different evidence rule. The 2024 figures were read through a summarising tool and want a re-check. 52167
Reported, not yet replicated
Where the list of 30 lives
Breuer et al. 2019, Supplementary file 3 (eLife, https://cdn.elifesciences.org/articles/36842/elife-36842-supp3-v2.xlsx): filter class 'Unclear' and essentiality 'Essential'. The loci (MMSYN1_ numbers) are 0033, 0143, 0145, 0146, 0249, 0281, 0286, 0296, 0317, 0325, 0338, 0352, 0373, 0379, 0388, 0412, 0424, 0430, 0431, 0439, 0440, 0478, 0511, 0636, 0691, 0778, 0797, 0805, 0835, 0906. Transporters are a separate class in that paper, so these are not mostly transporters. 2
Reported, not yet replicated
Interaction screens that exist
Yeast pairs: about 23 million double mutants, with about 550,000 negative and 350,000 positive interactions. Yeast triples: 195,666 triple mutants tested, 3,196 negative trigenic interactions. E. coli: about 600,000 pairs tested of about 8 million possible, with 25,239 aggravating and 17,466 alleviating interactions (163 chosen query genes); a second E. coli screen tested 315,400 double deletions and found 1,881 sick or lethal. No pair or triple map of syn3A, or of any Mycoplasma, was found. 11121314
Reported, not yet replicated
Physics inputs that were measured
GFP (27 kDa) diffuses at 87 um^2/s in dilute solution and at 3 to 14 um^2/s in E. coli cytoplasm (7.7 +/- 2.5 in one FRAP study); a 72 kDa fusion at 2.5, and a very large fusion is essentially immobile. Macromolecules fill 0.13 to 0.24 of the volume depending on growth conditions (a review's range; definitions of 'occupied' differ). A syn3A cell is a sphere of about 400 to 500 nm diameter. 15191
Reported, not yet replicated
Physics inputs that are guesses
The syn3A model sets every protein's diffusion coefficient to 1.0 um^2/s, borrowed from E. coli ribosome-assembly models; nobody measured it in a Mycoplasma. The review on prokaryote diffusion gives 1e8 to 1e9 per M per s for the fastest diffusion-limited association and 1e5 to 1e6 per M per s as a more realistic range for molecules with small binding patches; no direct determination of diffusion limitation in a living cell exists. 115
Reported, not yet replicated
Build it
Step 1. Count the parts: what would end to end mean?
End to end means every gene product and every reaction tracked through one full cell cycle. List what the best published model of the smallest free-living cell actually tracks, then list what it leaves out. 1
Commit first: how many molecule types and reactions does the published syn3A model track, and what is not in it?
7,765 molecule types, over 7,200 reactions, over 77,000 proteins, 503 ribosomes. Left out by the authors' own limitations list: regulatory proteins and riboswitches, base modifications, complex and ribosome assembly (folded into rates), a septum model for division, DNA supercoiling; one replication fork only; ribosomes and DNA stationary in the spatial run. A 2026 model reports one full cycle with growth, chromosome replication and division, but as an averaged model; only its abstract and press coverage were read.
Reported, not yet replicated
Step 2. Count the unknown genes
Take the genes of the cell and ask how many have no accepted function. The answer depends on what you accept as evidence, so name the class with every count. 21
Equation. fraction unknown = genes without accepted function / genes counted
Commit first: of syn3A's 452 protein genes, how many are of unclear function, and does 'about 70 to 90' hold?
91 of 452 in 2019 (30 essential, 32 quasi-essential, 29 non-essential) and 87 of 452 in 2022. So 70 to 90 sits at the edge of the syn3A count under one scheme, but no source defines a set of 70 to 90 genes. A 2026 computational recount puts 50 to 60 syn3.0 proteins without function, which is reannotation rather than new experiments.
Reported, not yet replicated
Step 3. Same genome, new rule
Divide unknown by total for E. coli K-12 under two published rules and compare. 67
Equation. 1,600 / 4,623 = 34.6% (2019); 738 / 4,747 = 15.5% (2024)
Commit first: did the unknown fraction fall because experiments filled in functions, or because the category was redefined?
The denominators differ (4,623 and 4,747), the 2024 scheme adds a middle 'partial' tier (1,247 genes) and sets some tiers by rules on the product name, and the paper does not list which genes moved. The two numbers are not one measurement repeated.
Our inference
Step 4. Argue it both ways: missing principle or backlog?
A missing principle predicts that characterised unknown essential genes turn out chemically new and that the unknown set resists every method. A backlog predicts that most resolve into known families once tested and that the hard cases are the technically hard ones, such as membrane proteins. Write what each predicts before looking. 829
Commit first: which observation would you accept as telling the two apart?
Nobody has run the decisive tally. In syn3.0, TMHMM predicts membrane helices in 49% of 'unknown' proteins, 35% of 'generic' and 6% of equivalogs, which fits a backlog of hard-to-assay proteins. Against that, the authors of the syn3A essential-metabolism paper say the unclear essential genes may point to new biological mechanisms, and four of the seven genes added to allow regular division are membrane-associated proteins of unknown function (abstract only). No bench-resolved function for a specific unknown gene was found in an abstract-level search.
Contested
Step 5. Function from sequence alone
Chain the inference: sequence search, structure prediction, structural match to a known family, then a function label. Ask what fraction of the hard class gets a specific label and how often such labels are right. 1020
Commit first: what share of the syn3.0 'unknown' proteins can a structure-based method give a specific molecular function?
C-I-TASSER assigned specific molecular-function terms to 86% of 438 syn3.0 proteins and 49% of the 'unknown' category: predictions, not bench results. No measured accuracy of structure-based function transfer on syn3A genes was found. Where function transfer by similarity was measured, in 37 enzyme families, misannotation was near 0% in the curated database for most families but 5% to 63% by superfamily in uncurated ones; those are well-studied enzymes, not bacterial unknowns.
Reported, not yet replicated
Step 6. Count the interaction space
Treat the protein genes as entities and count subsets. The number of k-gene combinations among n genes is the binomial coefficient. 11112
Equation. C(n,k) = n! / (k! (n-k)!); C(452,2) = 452*451/2; C(452,3) = 452*451*450/6
Commit first: roughly how many gene pairs and triples are there among 452 genes?
Pairs: 101,926. Triples: 15,288,900 (150 times the pairs). Quadruples: 1,716,179,025. Inputs: n = 452 (Thornburg 2022). The arithmetic is ours. For 473 genes (syn3.0): 111,628 pairs, 17,525,596 triples. The yeast screens tested about 23 million pairs and 195,666 triples; the triples are about 5 in a million of the roughly 4e10 possible (the 4e10 is our arithmetic). No pair map of syn3A exists, and higher orders are never sampled.
Our inference
Step 7. Measured density, and what it would imply
Real screens report the fraction of tested pairs that interact. Apply that density to the syn3A pair count as a thought experiment, not a prediction. 111314
Equation. expected interacting pairs = density x C(452,2)
Commit first: what range of interaction density do real screens give, and is it one number?
Yeast pairs: about 3.9% (900,000 called of about 23 million; our division). E. coli: 7.1% (42,705 of about 600,000) in one screen and 0.6% (1,881 of 315,400) in another, 12-fold apart in the same organism, with different media, query sets and thresholds. Applied to 101,926 syn3A pairs, 0.6% gives about 600 and 7.1% about 7,200: an extrapolation that assumes syn3A resembles E. coli screens and that the screened sets were representative, which they were not (queries were chosen genes). Nobody has measured it.
Our inference
Step 8. Let the molecules move: how long to cross the cell?
Brownian motion in three dimensions gives a mean squared displacement that grows linearly in time. Estimate the time to cross a syn3A-sized cell. 19151
Equation. <r^2> = 6 D t; t = L^2 / (6 D)
Commit first: how long does a typical protein take to cross a 0.4 um cell?
With L = 0.4 um: 2.7 ms at D = 10 um^2/s, 27 ms at D = 1 (the value the model uses) and 2.7 s at D = 0.01 (mRNA-like). Inputs for D are the E. coli range for GFP and the model's fixed value; the crossing times are our arithmetic and assume free three-dimensional diffusion with no obstacles. The measured GFP diffusion in E. coli is 7.7 +/- 2.5 um^2/s.
Our inference
Step 9. How long until two partners meet?
For two fully reactive spheres the diffusion-limited rate constant follows from Smoluchowski. Convert one molecule in a cell-sized volume to a concentration, then take the wait as the inverse of rate constant times concentration. 15
Equation. k = 4 pi (D1 + D2)(R1 + R2); c = 1 / (V N_A); t_wait = 1 / (k c)
Commit first: with a diffusion-limited rate, how often does one protein meet a single partner copy in a 400 to 500 nm cell?
Inputs: D1 = D2 = 10 um^2/s, R1 = R2 = 5 nm. Result k = 1.5e9 per M per s. A sphere of 400 to 500 nm diameter is 0.034 to 0.065 um^3, so one molecule is 25 to 50 nM. Wait: 13 to 27 ms at 1.5e9; 0.2 to 0.4 s at 1e8 (the review's quoted ceiling); 20 to 400 s at 1e5 to 1e6 (the review's realistic range). Our arithmetic. The review itself gives about 1 s for D = 10 in 1 um^3, with a simulation mean of 0.6 s; its 1e8 to 1e9 ceiling does not match the plain formula by a factor of about 15, and we could not reconcile this. Even the slowest wait is well under the 105-minute cycle for one pair; the unanswered question is how many such waits are chained.
Our inference
Step 10. Add crowding: does excluded volume alone explain the slowing?
Run a Brownian-dynamics cytoplasm with the real mix of macromolecules and see if GFP slows as measured. 161719
Commit first: if the cytoplasm is only hard-sphere crowding, does GFP diffuse as slowly as measured?
No. In one model (1,008 macromolecules of 51 types) steric crowding alone gave GFP diffusion 3 to 6 times too fast, and a fitted attraction term was needed to match. A second model (159 protein types, hydrodynamic interactions, no fitted stickiness) matched with GFP at 6.51 +/- 0.47 um^2/s versus 7.7 measured, and about 16 without hydrodynamics. Both reproduce one number with different ingredients, so agreement does not say which physics is responsible.
Contested
Where it breaks
End to end does not exist yet
The 2022 model tracks 20 minutes of the spatial cell with no growth or DNA replication, and its authors list what is left out: regulation, base modification, assembly, the division septum, supercoiling. The 2026 whole-cell model covers one cycle but averages individual molecules; its paper was not opened, so what it leaves out is not known to us. 1
Reported, not yet replicated
The model is partly fitted to what it reproduces
The calculated fraction of active RNA polymerase in the spatial run was 33%, but the well-stirred model uses 16% because it reproduces the measured doubling time. Earlier whole-cell models showed the same pattern: the E. coli model gave the wrong doubling time until literature values for polymerase and ribosome output were raised. A match to its own fitting target is a consistency check, not a test. No independent re-run or held-out benchmark was found. 143
Reported, not yet replicated
The count of unknown genes follows the rule
One E. coli genome goes from 34.6% to 15.5% unknown by changing the definition and adding a tier. Any count without its class definition is not comparable, including 149 for syn3.0 beside 91 for syn3A. 6752
Our inference
Function from sequence is a hypothesis for the hard class
Structure-based and sequence-based methods give labels to about half of the unknown class, but no measured accuracy on syn3A genes exists, fold is not function, membrane regions predict weakly, and similarity transfer has spread errors where it has been measured. For the 30 essential unclear genes, none has an experimentally established function in the sources read. 108202
Reported, not yet replicated
No interaction map for the cell itself
Every density comes from yeast, E. coli or other bacteria, and the same organism gives densities 12-fold apart depending on screen design. For the syn3A pair space (about 100,000) the triple space is 150 times larger, and higher orders have never been sampled in any organism. Redundant pairs, each non-essential alone and essential together, are why single-gene lists misled the first minimal design. 131411125
Reported, not yet replicated
The physics is imported, not derived
The model's protein diffusion coefficient is one borrowed number, fixed, and in the text we could search it is never varied; the 34% and 45% active-fraction outputs come from that single setting. Two crowding models reproduce one measurement with different physics, and one needed a fitted term. Diffusion tells you when molecules meet, not whether they react, and no cell reaction has been shown to be diffusion-limited in vivo. No diffusion coefficient has been measured in Mycoplasma or syn3A. 1161715
Reported, not yet replicated
What would settle it
Is an unknown essential gene a missing mechanism or a missing measurement?
Establish experimentally what one of the 30 essential unclear genes does, and ask whether its activity is chemically new or a known family. Most resolving into known families would support a backlog; a characterised gene with new chemistry would support a missing mechanism. No such tally has been run, and no bench-resolved function was found for any of them since 2019 (abstract-level search). Links: questions/unknown-genes-in-a-minimal-cell and questions/why-are-orphan-proteins-orphans. 29
Our inference
Is the whole-cell model predicting or fitting?
Pre-register predictions for quantities not used in tuning, measure them, and have a second group re-run the published code. Link: questions/one-model-to-rule-them-all. 14
Our inference
How much of the interaction space is real?
Build a pair map in syn3A or a related Mycoplasma. The gap search found single-guide CRISPRi in two mycoplasmas but no paired-guide tool and no cost estimate for a full library. Link: questions/has-every-protein-pair-been-tested. 1314
Our inference
What is the real diffusion coefficient inside a minimal cell?
Measure protein mobility in syn3A cytoplasm with probes of a size series, then repeat the model's protein diffusion sweep with the time step rescaled. Link: instruments/seeing-everything-in-a-cell-at-once, which is still a stub. 11915
Our inference
Try it in a running model
Teach this
In one sentence
Even the smallest living cell cannot yet be modelled end to end: about a fifth of its protein genes have no accepted function, and the best model borrows or fits several of its numbers. 21
Reported, not yet replicated
The five-minute version
Count first. The laboratory-made cell JCVI-syn3A has 493 genes; 452 make proteins, and 91 of those were of unclear function in 2019. The best published model tracks over 7,200 reactions for 20 minutes of cell time and leaves out regulation, assembly and division machinery; a 2026 model runs a whole cycle but averages the molecules. The unknown count depends on the rule: one E. coli genome goes from 34.6% to 15.5% unknown by redefining a category. Among 452 genes there are 101,926 pairs and 15,288,900 triples, and no map of even the pairs exists for this cell. If you only let molecules move, a protein crosses the cell in milliseconds and finds a partner in well under a minute; but physics alone mispredicts how fast proteins move in cytoplasm and says nothing about what reacts. 21671516
Reported, not yet replicated
What a listener will ask
Is 70 to 90 unknown genes right?
Does the unknown set prove a deep gap in biology?
Why not just simulate everything?
The published model leaves out regulation and division machinery, fits some rates to reproduce growth, and fixes diffusion at one borrowed value. 1
Reported, not yet replicated
The common wrong idea
A count of unknown genes is a measurement of how much biology we do not understand.
It is a count under an evidence rule; change the rule and the same genome goes from 34.6% to 15.5% unknown. 67
Reported, not yet replicated
The one citation to hand over: 1
Still open here
- Which of the 30 essential genes of unclear function in JCVI-syn3A has an experimentally established function, and is any of it chemically new? No bench result was found in an abstract-level search of 2019 to 2026.
- Is the 2026 whole-cell model of a full division cycle fitted to its targets or tested on held-out data? Only abstracts and press were read; the paper itself was not opened.
- What is the diffusion coefficient of a protein in syn3A cytoplasm, measured? None was found; the model borrows one number from E. coli work.
- How many of the roughly 100,000 gene pairs in syn3A interact? No pair map of any Mycoplasma was found.
- Does the exercise need a runnable piece on the page (a combinatorial counter and an encounter-time calculator) linked to the mounted minimal-cell-worlds tool?
Linked from
- Gene annotation, evidence codes and what 'unknown function' means
- Gene essentiality and the quasi-essential class
- Genetic interaction and the combinatorial space
- Macromolecular crowding and diffusion in bacterial cytoplasm
- JCVI-syn3A, the minimal synthetic cell
- Whole-cell model
- How many genes in the simplest cell do we still not understand, and what does that mean?
Sources
- Thornburg ZR, Bianchi DM, Brier TA, et al.. Fundamental behaviors emerge from simulations of a living minimal cell. Cell (2022). paper; opened 2026-11-17 all records citing this
- Breuer M, Earnest TM, Merryman C, et al.. Essential metabolism for a minimal cell. eLife (2019). paper; opened 2026-11-17 all records citing this
- Karr JR, Sanghvi JC, Macklin DN, et al.. A whole-cell computational model predicts phenotype from genotype. Cell 150(2) (2012). paper; opened 2026-11-17 all records citing this
- Macklin DN, Ahn-Horst TA, Choi H, et al.. Simultaneous cross-evaluation of heterogeneous E. coli datasets via mechanistic simulation. Science 369 (2020). paper; opened 2026-11-17 all records citing this
- Hutchison CA, Chuang RY, Noskov VN, et al.. Design and synthesis of a minimal bacterial genome. Science 351:aad6253 (2016). paper; opened 2026-11-17 all records citing this
- Ghatak S, King ZA, Sastry A, Palsson BO. The y-ome defines the 35% of Escherichia coli genes that lack experimental evidence of function. Nucleic Acids Research 47(5) (2019). paper; opened 2026-11-17 all records citing this
- Moore LR, Caspi R, Boyd D, et al., Karp PD. Revisiting the y-ome of Escherichia coli. Nucleic Acids Research (2024). paper; opened 2026-11-17 all records citing this
- Antczak M, Michaelis M, Wass MN. Environmental conditions shape the nature of a minimal bacterial genome. Nature Communications (2019). paper; opened 2026-11-17 all records citing this
- Pelletier JF, Sun L, Wise KS, et al.. Genetic requirements for cell division in a genomically minimal cell. Cell 184(9) (2021). paper; opened 2026-11-17 all records citing this
- Zhang C, Zheng W, Cheng M, Omenn GS, Freddolino PL, Zhang Y. Functions of Essential Genes and a Scale-Free Protein Interaction Network Revealed by Structure-Based Function and Interaction Prediction for a Minimal Genome. Journal of Proteome Research (2021). paper; opened 2026-11-17 all records citing this
- Costanzo M, VanderSluis B, Koch EN, et al.. A global genetic interaction network maps a wiring diagram of cellular function: supplementary data page. Boone Lab, University of Toronto (2016). organisation-page, not peer reviewed; opened 2026-11-17 all records citing this
- Kuzmin E, VanderSluis B, Wang W, et al.. Systematic analysis of complex genetic interactions: supplementary data page. Boone Lab, University of Toronto (2018). organisation-page, not peer reviewed; opened 2026-11-17 all records citing this
- Babu M, Arnold R, Bundalovic-Torma C, et al.. Quantitative genome-wide genetic interaction screens reveal global epistatic relationships of protein complexes in Escherichia coli. PLoS Genetics (2014). paper; opened 2026-11-17 all records citing this
- Cote JP, French S, Gehrke SS, et al.. The Genome-Wide Interaction Network of Nutrient Stress Genes in Escherichia coli. mBio 7(6) (2016). paper; opened 2026-11-17 all records citing this
- Schavemaker PE, Boersma AJ, Poolman B. How Important Is Protein Diffusion in Prokaryotes?. Frontiers in Molecular Biosciences 5 (2018). review; opened 2026-11-17 all records citing this
- McGuffee SR, Elcock AH. Diffusion, crowding & protein stability in a dynamic molecular model of the bacterial cytoplasm. PLoS Computational Biology (2010). paper; opened 2026-11-17 all records citing this
- Hasnain S, McClendon CL, Hsu MT, et al.. A new coarse-grained model for E. coli cytoplasm: accurate calculation of the diffusion coefficient of proteins and observation of anomalous diffusion. PLoS ONE 9(9) (2014). paper; opened 2026-11-17 all records citing this
- Fraser CM, Gocayne JD, White O, et al.. The minimal gene complement of Mycoplasma genitalium. Science 270(5235) (1995). paper; opened 2026-11-17 all records citing this
- Elowitz MB, Surette MG, Wolf PE, Stock JB, Leibler S. Protein mobility in the cytoplasm of Escherichia coli. Journal of Bacteriology 181(1) (1999). paper; opened 2026-11-17 all records citing this
- Schnoes AM, Brown SD, Dodevski I, Babbitt PC. Annotation error in public databases: misannotation of molecular function in enzyme superfamilies. PLoS Computational Biology (2009). paper; opened 2026-11-17 all records citing this
- Amit's words about the atlas, kept verbatim in this repository. Kept in the atlas repository; not published.
History of this record (8)
This record has been changed 8 times, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.
- Status
- partly written
- Last updated
- 2026-11-19
- Record
- content/exercises/model-a-single-bacterial-cell.json