MSScan.bin

Binary scan index file. Always present in .d/AcqData/. Encodes per-scan metadata and byte offsets into the payload files (MSPeak.bin / MSProfile.bin).

Status: Confirmed

File Layout

FileSize = GlobalHeaderSize + NumOfScans * RecordStride holds exactly for every tested file with zero remainder.

The GlobalHeaderSize is stored as a 32-bit little-endian integer at file offset 0x58 (88).

Validated corpus (8 files, all with zero remainder):

PXD	Instrument	Model	HeaderSize	Stride
PXD004747	QQQ (MRM)	G6410A	296	196
PXD028295	QQQ (MRM)	G6410A	296	186
PXD001310	Q-TOF (Profile+Peak)	G6550A	228	284
PXD004426	Q-TOF (Peak only)	G6550A	228	220
PXD007734	Q-TOF (Peak only)	G6540B	228	220
PXD011212	Q-TOF (Peak only)	G6550A	228	220
PXD014285	Q-TOF (Peak only)	G6550A	228	220
PXD031771	Q-TOF (Peak only)	-	228	216

Instrument models were read from Devices.xml. G6540B and G6550A are both Q-TOF instruments; G6410A is a triple-quadrupole (QQQ). AcqSoftwareVersion in Contents.xml confirms the 6200 series TOF/6500 series Q-TOF firmware label on all Q-TOF files. The two distinct QQQ strides (196 vs 186) are confirmed from two separate G6410A datasets with different run configurations. Stride 216 was found during full-corpus conformance testing and required disambiguating from stride 196 (see the stride-detection note below).

MSProfile.bin files that exist but are 0 bytes (PXD011212, PXD014285) are placeholders; their scan records have no profile SpectrumParams block.

Global Header

Little-endian. Starts with 0x01 0x01 0x00 0x00 (value 257 as int32). The GlobalHeaderSize field at byte offset 0x58 is the primary structural anchor.

Scan Record Fields (Confirmed)

All fields are little-endian. Offsets are relative to the start of each scan record (i.e., after the global header).

Field confirmed at record offset 0 for all files:

• ScanID (int32): 1-based sequential scan number.

Field confirmed at record offset 4:

• QQQ: MRM channel/compound ID (int32). Cycles through the defined MRM transitions per acquisition cycle.
• Q-TOF: unknown small int (often 1).

Field confirmed at record offset 12 for all files:

• RetentionTime (double, minutes): Verified monotonically increasing (zero inversions) across all scans in PXD004747 (0.0001-20.3 min), PXD028295 (2.35-41.19 min), PXD001310 (0.038-60.13 min). For QQQ MRM, all scans within an MRM cycle share the same chromatographic retention time (the MRM channels all fire at the same chromatographic point).

Field confirmed at record offset 20:

• MSLevel (int16, not int32 - the upper two bytes at offset 22 belong to a separate field and were previously misread as part of MSLevel): 1 for MS1, 2 for MS2. Observed in Q-TOF (PXD001310) where scan 8 had MSLevel=2 and had a precursor ion listed at other offsets.

Q-TOF-specific fields (confirmed for stride=284):

• Offset 36: TIC or related intensity metric (double).
• Offset 44: Related intensity metric (double).
• Offset 84: Precursor Target m/z (f64, 8 bytes). Confirmed for MS2 records in Q-TOF data across multiple datasets (e.g. PXD004426, PXD007734, PXD001310). Represents the center of the isolation window for Auto-MS/MS scans.
• Offset 244: MinX - minimum m/z of the scan window (double, Da). Verified equal to the scan's declared mass range lower bound.
• Offset 252: MaxX - maximum m/z of the scan window (double, Da).

SpectrumParamsType Block (Confirmed)

Each scan record contains one or two SpectrumParamsType blocks. For files with both profile and centroid data (PXD001310), a block with SpectrumFormatID=1 appears at record offset 156 and a block with SpectrumFormatID=2 appears at record offset 220.

SpectrumFormatID Values (Verified)

These IDs are NOT run-level properties. They are per-scan and per-block:

• ID = 1: Profile-mode data, pointing into MSProfile.bin. Only observed in PXD001310 (Q-TOF with profile mode enabled). Stores SpectrumOffset (int64), ByteCount (int32, compressed size), PointCount (int32, uncompressed point count), UncompressedByteCount (int32).
• ID = 2: Centroid peak data, pointing into MSPeak.bin. Observed in all Q-TOF files. Does NOT have an UncompressedByteCount field (data is not compressed).
• ID = 3: Single-point MRM data, pointing into MSPeak.bin. Observed in both QQQ files (PXD004747, PXD028295). Structurally same as ID=2 but always has PointCount=1.

Note on ID=3 encoding: In PXD028295 (stride=186), the SpectrumFormatID is stored as int16 (not int32) at the params block offset, with the SpectrumOffset (int64) immediately following at +2 bytes. In all other files (stride >= 220), the SpectrumFormatID is int32 and SpectrumOffset starts at +4 bytes. This is the only structural divergence found across the corpus.

SpectrumParams Block Offsets Within Record

Stride	First block offset	FormatID width
186	136	int16
196	144	int32
216	152	int32
220	156	int32
284	156 (profile) + 220 (centroid)	int32

Block Field Layout (FormatID as int32)

Bytes	Field	Type
0-3	SpectrumFormatID	int32
4-11	SpectrumOffset	int64
12-15	ByteCount	int32
16-19	PointCount	int32
20-23	UncompressedByteCount	int32 (only present when FormatID=1)

Stride Detection

Since multiple candidate strides can evenly divide a file's payload length (e.g. 216 and 196 bytes both divide some Q-TOF payloads), picking the first evenly-dividing candidate is not sufficient. The reader validates a candidate stride by checking that the second record, decoded at that stride, has a plausible MSLevel (1 or 2) and a ScanID that strictly increases from the first record - only then is the candidate accepted.

Unresolved

• The ScanID field in PXD028295 starts at 1116 instead of 1, suggesting MSScan.bin may represent only a portion of a longer run, or scan numbering is session-level rather than file-level.
• Full field map for the bytes between RetentionTime (offset 12) and the first confirmed MSLevel field or mass-range fields, aside from the now-confirmed offsets 36, 44, and 84.
• The source of the 254 SpectrumOffset inversions in PXD028295 is not yet explained. The inversions appear at MRM channel boundaries rather than being random, which may indicate that within a multi-channel MRM cycle some channels are written to MSPeak.bin out of scan-number order. This is a known consequence of interleaved multi-channel MRM acquisition rather than a format error.