Extracted from MDL CTFile Formats (October 2003) + BIOVIA 2020 additions. Focused on molecular structure + stereo information relevant to chython.

A molfile = Header Block + Connection Table (CTAB).

Line 1: Molecule name (max 80 chars, unformatted)
Line 2: IIPPPPPPPPMMDDYYHHmmddSSssssssssssEEEEEEEEEEEERRRRRR
         (initials, program, date, dimensions, scaling, energy, registry)
Line 3: Comment (blank if none)

Lines 2-3 may be blank. Line 1 must NOT start with $MDL, $$$$, $RXN, or $RDFILE.

aaabbblllfffcccsssxxxrrrpppiiimmmvvvvvv

One line per atom, fixed-width columns:

xxxxx.xxxxyyyyy.yyyyzzzzz.zzzz aaaddcccssshhhbbbvvvHHHrrriiimmmnnneee

Important: Fields dd and ccc are superseded by M ISO, M CHG, M RAD in properties block.

One line per bond, fixed-width columns:

111222tttsssxxxrrrccc

Stereo convention: The wedge (pointed) end is at the FIRST atom (field 111).

Terminated by M END. Key properties for molecules:

M  CHGnn8 aaa vvv ...       Charge: vvv = -15..+15
M  RADnn8 aaa vvv ...       Radical: 0=none, 1=singlet, 2=doublet, 3=triplet
M  ISOnn8 aaa vvv ...       Isotope: absolute atomic mass (positive integer)
M  END                      End of CTAB

When M CHG/M RAD present, they supersede ALL atom block charge/radical values (forces 0 on unlisted atoms).

Not in the 2003 spec. Uses Sgroup-like property lines:

M  STY  1   1 DAT           Define Sgroup 1 as data type
M  SAL   1  n  a1 a2 ...    Atoms in stereo group
M  SDT   1  MDLV30/STERAC1  Field name identifies stereo type
M  SED   1                   (empty data)

Types: MDLV30/STEABS, MDLV30/STERAC{n}, MDLV30/STEREL{n} (same semantics as V3000).

V3000 file = V2000 "no structure" header (with version stamp V3000) + extended CTAB blocks.

• Every line begins with M V30 (2 spaces after M, 1 after 30)
• Line continuation: - as last char, next line's M V30 prefix stripped and concatenated
• Max 80 chars per physical line
• Values: positional first, then KEYWORD=value optional
• List values: KEYWORD=(N val1 val2 ... valN) where N = count
• Strings with spaces/parens/quotes must be double-quoted; literal " doubled

{V2000 header: name, program line, comment, counts "0  0  0     0  0            999 V3000"}
M  V30 BEGIN CTAB
M  V30 COUNTS na nb nsg n3d chiral
M  V30 BEGIN ATOM
...atoms...
M  V30 END ATOM
M  V30 BEGIN BOND
...bonds...
M  V30 END BOND
[M  V30 BEGIN SGROUP ... M  V30 END SGROUP]
[M  V30 BEGIN COLLECTION ... M  V30 END COLLECTION]
M  V30 END CTAB
M  END

M  V30 COUNTS na nb nsg n3d chiral [REGNO=regno]

M  V30 BEGIN ATOM
M  V30 index type x y z aamap [CHG=val] [RAD=val] [CFG=val] [MASS=val] [VAL=val] ...
M  V30 END ATOM

Query-only (ignore): HCOUNT, STBOX, SUBST, UNSAT, RBCNT, ATTCHPT, RGROUPS, ATTCHORD. Reaction-only: INVRET (0/1/2), EXACHG (0/1).

M  V30 BEGIN BOND
M  V30 index type atom1 atom2 [CFG=val] [TOPO=val] [RXCTR=val] [STBOX=val]
M  V30 END BOND

Note V3000 vs V2000 bond stereo difference: V2000 uses 1=Up, 6=Down. V3000 uses 1=Up, 3=Down.

M  V30 BEGIN COLLECTION
M  V30 MDLV30/STEABS ATOMS=(n a1 a2 ...)
M  V30 MDLV30/STERAC1 ATOMS=(n a1 a2 ...)
M  V30 MDLV30/STEREL1 ATOMS=(n a1 a2 ...)
M  V30 END COLLECTION

• n is an integer >= 1 identifying the group
• Multiple groups can coexist (e.g., STERAC1, STERAC2, STEREL1)
• Atoms not listed in any collection default to ABS
• ATOMS list format: (count atom_index atom_index ...)

M  V30 BEGIN SGROUP
M  V30 index type extindex [ATOMS=(...)] [FIELDNAME=name] [FIELDDATA=data] ...
M  V30 END SGROUP

Types: SUP(eratom), MUL(tiple), SRU, MON(omer), COP(olymer), DAT(a), etc.

Only relevant for chython: DAT type SGROUPs with FIELDNAME="MRV_IMPLICIT_H" and FIELDDATA specifying implicit H count.

Multiple molecules + associated data. Format:

[Molfile]            <- V2000 or V3000 MOL block (header + CTAB + M END)
>  <FIELD_NAME>      <- Data header (field name in angle brackets)
data value           <- One or more lines of data
                     <- Blank line terminates data item
>  <ANOTHER_FIELD>   <- Repeat for each data field
more data

$$$$                 <- Record delimiter (separates molecules)

> [registry_info] <field_name>

The > must be in column 1. Field name is in <angle_brackets>.

Check line 4 (counts line): if it starts with M V30 BEGIN CTAB then V3000, else V2000. (Line indices: 0=name, 1=program, 2=comment, 3=counts/begin, 4=first CTAB line for V3000)

Actually, the standard detection: check if the version stamp in the counts line (chars 34-39) is V3000. Or pragmatically: line[4] (0-indexed from start of MOL block) starting with M V30.

$RXN
reaction name
      IIIIIIPPPPPPPPPMMDDYYYYHHmmRRRRRRR    (program info line)
comment line
rrrppp                                       (counts: 3-digit reactants, 3-digit products)
$MOL
[Molfile for reactant 1]
$MOL
[Molfile for reactant 2]
...
$MOL
[Molfile for product 1]
...

• Line 1: $RXN identifier
• Line 2: Reaction name (or blank)
• Line 3: Program info (or blank)
• Line 4: Comment (or blank)
• Line 5: rrrppp - number of reactants (3 chars) + number of products (3 chars)
  • Extended: rrrpppaaa with optional agent count
• Each $MOL delimiter followed by a complete Molfile (header + CTAB + M END)
• Order: all reactants first, then all products, then agents (if any)

$RXN V3000
reaction name
program info
comment
M  V30 COUNTS nreactants nproducts [nagents]
M  V30 BEGIN REACTANT
M  V30 BEGIN CTAB
...ctab for reactant 1...
M  V30 END CTAB
M  V30 BEGIN CTAB
...ctab for reactant 2...
M  V30 END CTAB
M  V30 END REACTANT
M  V30 BEGIN PRODUCT
M  V30 BEGIN CTAB
...ctab for product 1...
M  V30 END CTAB
M  V30 END PRODUCT
[M  V30 BEGIN AGENT
M  V30 BEGIN CTAB...END CTAB
M  V30 END AGENT]
M  END

• Line 1: $RXN V3000 (the V3000 token distinguishes from V2000)
• Lines 2-4: name, program info, comment (same as V2000 rxn header)
• Counts line: M V30 COUNTS nreactants nproducts [nagents]
• Each molecule is a full CTAB block (same format as V3000 MOL, without the outer header)
• No $MOL delimiters — molecules are wrapped in BEGIN/END CTAB pairs
• No per-molecule headers (no name/program/comment per molecule)

Contains molecules OR reactions with associated data. More general than SDF.

$RDFILE 1                        <- Header (required, file start)
$DATM    MM/DD/YY HH:mm         <- Date stamp (treated as comment)
$RFMT [$RIREG regno]            <- Reaction record start (or $MFMT for molecule)
$RXN                             <- Embedded rxnfile
...rxnfile content...
M  END
$DTYPE field_name                <- Data field identifier
$DATUM data_value                <- Data value
$DTYPE another_field
$DATUM another_value
$RFMT                            <- Next record starts here
...

• $MFMT — molecule record (followed by embedded molfile)
• $RFMT — reaction record (followed by embedded rxnfile starting with $RXN)
• $MIREG / $RIREG — internal registry reference
• $MEREG / $REREG — external registry reference

• $DTYPE field_name — field name (one per data item)
• $DATUM value — data value (can be multi-line for fields >80 chars)

• No $$$$ delimiter (records separated by next $RFMT/$MFMT)
• Can contain both molecules and reactions in same file
• Uses $DTYPE/$DATUM instead of > <field>
• Has file-level header ($RDFILE, $DATM)
• No blank lines allowed except within embedded mol/rxn blocks

For single bonds at tetrahedral centers:

• Up (wedge): bond goes from center atom (first atom) ABOVE the plane toward second atom
• Down (hash/dash): bond goes from center atom BELOW the plane toward second atom
• The pointed end of the wedge is at the first atom in the bond definition

For double bonds (cis/trans):

• Stereo determined from 2D coordinates of substituents
• Value 0 = use coordinates; value 3 = either (ignore stereo)

Calculated by viewing the center from behind the highest-numbered neighbor:

• 1 = odd parity: atoms 1,2,3 in clockwise order
• 2 = even parity: atoms 1,2,3 in counterclockwise order
• 3 = either: unmarked or racemic at that center

Note: In V2000, the sss field in atom block is IGNORED by readers. Stereo is determined from bond wedges + coordinates. In V3000, atom CFG is informational; bond CFG (wedge/hash) is what defines stereo.

Defines groups of stereocenters with specific epistemic relationships:

Group numbering allows multiple independent groups:

• STERAC1, STERAC2 = two independent AND groups (flip independently)
• STEREL1, STEREL2 = two independent OR groups

Within a single group, all atoms flip together (their relative configuration is fixed).

Star atoms (*) represent attachment points with multiple possible endpoints:

M  V30 5 1 1 7            <- bond from atom 1 to star atom 7
                              with ENDPTS on the bond or star atom
M  V30 5 1 1 7 ENDPTS=(3 2 3 4)   <- star atom 7 connects to atoms 2, 3, or 4

Format: ENDPTS=(N atom1 atom2 ... atomN) where N is count of endpoint atoms.

In practice: create a bond (type 8/special) from the real atom to each endpoint atom.