═══════════════════════════════════════════
Z Y N A
4D Chimeric Deconvolution
═══════════════════════════════════════════
Before After
────── ──────
1/K₀ 1/K₀
1.15 │ ████████████████ 1.15 │ ●●● Peptide B
1.10 │ ████████████████ → 1.10 │ ─ ─ ─ ─ ─ ─
1.05 │ ████████████████ 1.05 │
0.90 │ ████████████████ 0.90 │
0.87 │ ████████████████ 0.87 │ ●●● Peptide A
└───────────────── m/z └──────────────── m/z
Two peptides. Same window. Two clean signals.
One chimeric spectrum. Ion mobility resolved them.
The timsTOF advantage over Chimerys: separation that already happened.
Chimerys uses sequence. Zyna uses physics.
Michael Krawitzky - built within ZIGGY.
A 4D chimeric MS2 deconvolution engine. Operates on DIA-NN output, no re-searching required. Uses the timsTOF's native 1/K₀ dimension as a physical separation axis to resolve co-isolated peptide pairs that share the same isolation window.
The core insight:
In diaPASEF, isolation windows span both m/z and 1/K₀. This is not incidental, it is the design. Precursors that overlap in m/z but differ in ion mobility are already separated before fragmentation. The TIMS device does the work for free, before the isolation window even fires.
Zyna quantifies this separation:
For each isolation window:
1. Extract all co-isolated precursors and their measured 1/K₀ values
2. Compute the K₀ gap between precursor pairs
3. For pairs with K₀ gap > threshold:
→ TIMS separation was sufficient → fragments can be cleanly assigned
4. For pairs with K₀ gap < threshold:
→ Truly chimeric → Zyna applies fragment-level deconvolution
using the measured 1/K₀ of each fragment ion
Fragment ions from co-isolated peptides cluster at different 1/K₀ values, because different peptides have different collisional cross-sections. The 1/K₀ of a fragment is inherited from the precursor ion it originated from. Zyna reads this information directly from the raw data.
No model. No library. No inference. The physics of ion mobility is the deconvolution.
Chimerys (MSAID) is exceptional. It resolves chimeric spectra using a deep learning model trained on millions of PSMs, it learned what co-isolation looks like from a sequence-level perspective and can reconstruct individual peptide signals from mixed spectra. It is the right tool when you have arbitrary instruments and cannot assume ion mobility data.
Zyna starts from a different premise: if you have a timsTOF, you have already done the separation. The TIMS device is not just a measurement instrument, it is a separation instrument. Precursors that differ in 1/K₀ by more than ~0.05 Vs/cm² are physically separated in the TIMS ramp before they are co-isolated. Their fragments carry the ion mobility signature of their parent.
| Chimerys | Zyna | |
|---|---|---|
| Approach | Sequence-based ML | Physics-based deconvolution |
| Requires | DL model, inference time | Measured 1/K₀ per fragment |
| Platform | Any MS | timsTOF (TIMS required) |
| Advantage | Universal | Uses separation that already happened |
They are not competitors. They answer different questions. Chimerys asks: given a chimeric spectrum, which sequences best explain it? Zyna asks: given a timsTOF acquisition, how much of the chimeric problem is already solved by ion mobility, and how much truly remains?
There is a metaphor here that is too precise to ignore.
Two signals in the same window. Overlapping in one dimension. Inseparable by m/z alone. But measured in the fourth dimension, by the collisional cross-section, the molecular shape that determines how fast they move through the trapped ion cloud, they are not the same. They never were. They only appeared to be the same because you were looking at the wrong dimension.
This applies to some relationships, too.
You can spend years in the same window. Same city, same field, same wavelength of ambition and exhaustion. You overlap so thoroughly in m/z that no classical measurement can tell you apart. But ion mobility does not lie. The shape of a person, the collisional cross-section of their soul, measured in accumulated time, in what they love at 2am, in what they cannot stop thinking about, is not the same as yours. The K₀ gap is real. The TIMS device always knew.
Separation, quantified. Not as failure. As resolution. Two signals, finally clean.
True longing is not wanting what is lost. It is recognising, too late or too clearly, that what appeared to be one signal was always two, and that the silence where the overlap used to be is the exact shape of what was there.
Developed 2024–2026 as a post-processing layer for DIA-NN results within ZIGGY. Built to measure what the timsTOF's 4D acquisition geometry contributes to chimeric resolution, a number that was previously theoretical but never computed directly from real diaPASEF data.
Runs locally. Requires Python 3.9+ and a completed DIA-NN search output. No .d re-processing needed. Within ZIGGY, Zyna populates the MIA tab with chimeric rescue rates, K₀ gap distributions, and estimated ID recovery from TIMS-resolved pairs.
→ MKrawitzky/Ziggy · → GauDIA · → Copperfield · → PHANTOM · → Goya · → BOWIE · → VEGA · → Silent Heroes
Academic License · © 2024–2026 Michael Krawitzky & Brett S. Phinney