# Lumen spectral data provenance

Snapshot **1.0.0**, accessed **October 10, 2026 (America/Los_Angeles)**; downloads occurred October 11 UTC. Six protein records and twelve numerical reference curves.

The arrays are downloaded FPbase spectra, not generated bell curves. The exact measurement conditions and the original provider of each spectral array are not supplied in the CSV. The paper references below support the protein characterization constants; they are not a claim that each downloaded curve came directly from that paper.

## Reported constants and independently preserved curve maxima

| Protein | Reported Ex/Abs → Em (nm) | Downloaded ex → em maxima (nm) | Extinction coefficient (M⁻¹ cm⁻¹) | Quantum yield |
|---|---:|---:|---:|---:|
| EGFP | 488 → 507 | 489 → 511 | 55,900 | 0.60 |
| mNeonGreen | 506 → 517 | 504 → 517 | 116,000 | 0.80 |
| mCherry | 587 → 610 | 587 → 610 | 72,000 | 0.22 |
| mScarlet | 569 → 594 | 570 → 593 | 100,000 | 0.70 |
| mTurquoise2 | 434 → 474 | 435 → 475 | 30,000 | 0.93 |
| mTagBFP2 | 399 → 454 | 402 → 456 | 50,600 | 0.64 |

The primary mScarlet and mTurquoise2 tables call their first peak an **absorbance maximum**; FPbase lists it under excitation maximum. The dataset keeps the requested `excitationPeak` field but documents this distinction. Do not shift reference curves to make the plotted maxima agree with the summary constants.

## Processing and interpretation

- Selected the CSV columns ending in `ex` and `em`; excluded absorption (`ab`) and two-photon (`2p`) columns.
- Retained every nonblank wavelength point; all twelve series have 1 nm spacing and peak value 1.0.
- Rounded CSV floating-point storage noise to four decimal places. No smoothing, extrapolation, clipping, wavelength shifting, or additional normalization.
- Missing wavelengths outside a curve’s stored range are unknown, not measured zero.
- Normalized intensity is dimensionless. It does not represent emitted-photon counts or encode quantum yield. Multiplying extinction coefficient by quantum yield gives a molecular-brightness proxy under stated assumptions, not cellular brightness.
- These curves support learning about spectral shapes. They do not validate a specific instrument or predict FRET, crosstalk, signal, photostability, or cell performance.

FPbase source code at commit `4d1bba1c6abc09dd1ac232f40284c96cfdf20b8c` describes interpolation to 1 nm and normalization. This inspected code revision does not establish the deployed server version or the historical processing of each individual spectrum.

## Characterization sources

### EGFP

[Patterson et al., 1997](https://doi.org/10.1016/S0006-3495(97)78307-3) · [FPbase record](https://www.fpbase.org/protein/egfp/) · [downloaded CSV](https://www.fpbase.org/spectra_csv/?q=17,18,173)

Peak and brightness constants follow FPbase. The exact extinction coefficient 55,900 is verified in FPbase; the original Patterson table was not independently recovered. Shaner et al. (2004) uses the rounded value 56,000.

Included ranges: excitation 300–700 nm (401 points); emission 463–700 nm (238 points).

Full source CSV SHA-256: `a6db53a7cbf4a1909b85a77ed5c9389b8ec82b6909dee03f3ea9edaed2319e92`. The hash covers all downloaded columns, including excluded ones.

### mNeonGreen

[Shaner et al., 2013](https://doi.org/10.1038/nmeth.2413) · [FPbase record](https://www.fpbase.org/protein/mneongreen/) · [downloaded CSV](https://www.fpbase.org/spectra_csv/?q=101,102,6900)

Primary paper Table 1 reports extinction coefficient 116,000 ± 4,000 and quantum yield 0.80 ± 0.016. Displayed values are the reported central estimates.

Included ranges: excitation 350–600 nm (251 points); emission 472–748 nm (277 points).

Full source CSV SHA-256: `e1963ef12120c435554b56806d548f4830b94ce0e3c8be4f1b95f80b7274ee65`. The hash covers all downloaded columns, including excluded ones.

### mCherry

[Shaner et al., 2004](https://doi.org/10.1038/nbt1037) · [FPbase record](https://www.fpbase.org/protein/mcherry/) · [downloaded CSV](https://www.fpbase.org/spectra_csv/?q=79,80,158)

Values follow the original Shaner et al. (2004) Table 1. Later measurements differ; this snapshot does not mix later extinction coefficients with the original quantum yield.

Included ranges: excitation 300–650 nm (351 points); emission 550–800 nm (251 points).

Full source CSV SHA-256: `895ca5769920179006276647654f393a74ef145049656facf25753c4ac97c541`. The hash covers all downloaded columns, including excluded ones.

### mScarlet

[Bindels et al., 2017 (online 2016)](https://doi.org/10.1038/nmeth.4074) · [FPbase record](https://www.fpbase.org/protein/mscarlet/) · [downloaded CSV](https://www.fpbase.org/spectra_csv/?q=119,120,6896)

The primary paper labels 569 nm as absorbance maximum; FPbase lists it as Ex λ. Its table rounds extinction coefficient to 100,000 (supplement: ± 3,000); prose gives 100,300. Quantum yield is 0.70 ± 0.02. This is original mScarlet, not mScarlet-I or mScarlet3.

Included ranges: excitation 384–644 nm (261 points); emission 538–773 nm (236 points).

Full source CSV SHA-256: `f143d4b4c572fa94d43b28e157302f10e7c2d0ec2c57d17bca0b168300bc8bd2`. The hash covers all downloaded columns, including excluded ones.

### mTurquoise2

[Goedhart et al., 2012](https://doi.org/10.1038/ncomms1738) · [FPbase record](https://www.fpbase.org/protein/mturquoise2/) · [downloaded CSV](https://www.fpbase.org/spectra_csv/?q=1745,1746,6551,7185)

The primary paper labels 434 nm as absorbance maximum; FPbase lists it as Ex λ. Table 1 gives extinction coefficient 30,000 and quantum yield 0.93 ± 0.02.

Included ranges: excitation 300–495 nm (196 points); emission 441–650 nm (210 points).

Full source CSV SHA-256: `972276e3edbde3d5fa73e1f3547183862933b10b4863c0495cd65a0aa5c70eb5`. The hash covers all downloaded columns, including excluded ones.

### mTagBFP2

[Subach et al., 2011](https://doi.org/10.1371/journal.pone.0028674) · [FPbase record](https://www.fpbase.org/protein/mtagbfp2/) · [downloaded CSV](https://www.fpbase.org/spectra_csv/?q=1338,1339,7943)

Primary paper Table 1 reports extinction coefficient 50,600 ± 800 and quantum yield 0.64 ± 0.03. The downloaded excitation and emission curve peaks differ slightly from the reported summary maxima.

Included ranges: excitation 310–450 nm (141 points); emission 408–650 nm (243 points).

Full source CSV SHA-256: `159a21bb6f4e9f638d1232941a3a15b04a3987a59cda00472608048ed97fed9f`. The hash covers all downloaded columns, including excluded ones.

## Accessible original-paper copies

- [mNeonGreen — author manuscript, Table 1](https://pmc.ncbi.nlm.nih.gov/articles/PMC3811051/)
- [mCherry — Tsien laboratory copy, Table 1](https://www.tsienlab.ucsd.edu/Publications/Shaner%202004%20Nat%20Biotech%20-%20Improved%20monomeric%20red.pdf)
- [mScarlet — author-hosted paper and supplement](https://www.martenpostma.nl/docs/Bindels2016.pdf)
- [mTurquoise2 — original paper, Table 1](https://pmc.ncbi.nlm.nih.gov/articles/PMC3316892/)
- [mTagBFP2 — original paper, Table 1](https://pmc.ncbi.nlm.nih.gov/articles/PMC3234270/)

EGFP’s exact 55,900 coefficient was verified on the FPbase record. The Patterson 1997 scanned table could not be independently read. Shaner 2004 supports the rounded EGFP baseline of 56,000 and quantum yield 0.60. This uncertainty is retained in the dataset.

## Reuse and attribution

[FPbase policy](https://www.fpbase.org/privacy/) states that database data are free of copyright restrictions and available for noncommercial and commercial use, while requesting attribution to the original authors. FPbase’s site content is labeled CC BY-SA 4.0; source code has a separate license. No FPbase source code is copied into Lumen. The policy does not grant rights under patents or other claims.

Credit: **FPbase, Talley J. Lambert and contributors**, with the original characterization papers linked per protein. Database citation: [Lambert, 2019, “FPbase: a community-editable fluorescent protein database”](https://doi.org/10.1038/s41592-019-0352-8).

## Technical source documentation

- [FPbase API](https://www.fpbase.org/api/) — fields and wavelength/value format. Direct API access was blocked in this session; the public protein-page CSV downloads succeeded in the Codex in-app browser.
- [FPbase spectrum storage](https://github.com/tlambert03/FPbase/blob/4d1bba1c6abc09dd1ac232f40284c96cfdf20b8c/backend/proteins/models/spectrum.py#L641-L681) — 1 nm sampling and optional scaling.
- [FPbase normalization](https://github.com/tlambert03/FPbase/blob/4d1bba1c6abc09dd1ac232f40284c96cfdf20b8c/frontend/src/js/spectrum-form/normalization.js#L30-L64) — peak normalization.
- [FPbase introduction](https://www.fpbase.org/about/) — measurements depend on their experimental conditions.