Phylogenetic Dating Tree Error: Error: Cannot read the file named "*.timetree.subst"

[tahirsajid]

I am using IQ-TREE2 to construct molecular clock tree from H5N1 virus whole genome data with the following command:
iqtree2 -s ${TEMP}Combined_AllSegments.phy -p ${TEMP}AllSegments_Coordinates_${BATCH}.map -m MFP+MERGE --date ${INFO}Sample_Time_phyFile_Updated.txt -o "S000000001" --keep-ident -B 1000 -T AUTO -ntmax 5 --prefix ${PHYL}WGS_MolClock${BATCH}

The input alignment file is in Phylip ID format

The date file (Sample_Time_phyFile_Updated.txt) is like (first three lines):
S000000001 2025-10-20
S000000002 2026-06-15
S000000003 2026-07-08

The analysis run up to generation of maximum likelihood tree and saves the files, but when it proceeds for phylogenetic dating analysis, it produces the error:
"Error: Cannot read the file named "/Analysis/Phylogenetic_Tree/WGS_MolClock17-08-2026.timetree.subst""

I checked my results directory but there is no file with suffix * timetree.subst or *.timetree.outgroup

I am attaching the log file here. If somebody can advise, where the problem is coming from and how to fix it, it will be highly appreciated.
WGS_MolClock17-08-2026.log

[roblanf]

Hi @tahirsajid,

I had Claude look into it, and here's what it thinks. It looks correct to me. Please respond here to tell me if it worked.

Rob

From Claude...

The missing .timetree.subst and .timetree.outgroup files are a red herring — they're supposed to be missing. IQ-TREE embeds LSD2 as a library and passes the tree, outgroup and dates via in-memory streams, not files. Those filenames are just placeholders handed to LSD2's argument parser (the source literally comments them as // only fake file), and they're written to disk only under --date-debug.

The actual problem is your build. LSD2's option parser contains this check:

case 'i':
#ifndef USE_LSD2
    if( access( optarg, R_OK )!=0 )
        myExit( "Cannot read the file named \"%s\"\n", optarg );
#endif

That check is compiled out when USE_LSD2 is defined — exactly because the embedded path uses placeholder names. In your binary, LSD2 was compiled without -DUSE_LSD2, so it re-arms the check and aborts on a file IQ-TREE never intended to create. It dies during argument parsing, before any dating work happens, which is why the ML stage completes normally and you get nothing else.

This has come up twice before with the same error and the same cause: #61 (resolved by the reporter rebuilding) and #107, where @tothuhien suggested the workaround below.

Immediate workaround — add --date-debug to your existing command:

iqtree2 -s ... -p ... --date ... -o "S000000001" --date-debug --prefix ...

This materialises the three files on disk before LSD2 is invoked, so the access() check passes and dating proceeds normally.

Please also move to IQ-TREE 3. We'd encourage all new analyses to use IQ-TREE 3 — please grab the latest release, v3.1.3. Your command line works there essentially unchanged (just iqtree3 instead of iqtree2).

If you'd rather build from source, make sure the submodule is initialised so the definition actually reaches LSD2:

git submodule update --init lsd2
cmake -DUSE_LSD2=ON ..

A couple of things that are not the cause, in case you were suspecting them: the hyphens in your --prefix are harmless (getopt only treats an argument as an option if it begins with -, and yours begins with /), and neither -p nor -o is a problem — partitioned dating is fully supported.

One thing worth double-checking while you're at it: your date file must be YYYY-MM-DD. The sample rows you posted look right, but your ${BATCH} naming is DD-MM-YYYY, and a DD-MM-YYYY date file would be accepted silently by IQ-TREE and then misread by LSD2 — no error, just wrong dates. Also, keep --keep-ident: without it, deduplication can drop sequences that your date file references, which throws a different error.

— drafted by Claude Opus 5

[tahirsajid]

Thanks, @roblanf.
Addition of --date-debug in the command line worked for this time.
I will get new version installed.

[roblanf]

Just one thing I've noticed here too - it looks like you might be concatenating across segments. Just wanting to double check that you've made sure that's a viable thing to do here (i.e. that you're sure there's zero reassortment in your sequences).

Rob

[tahirsajid]

Thanks for highlighting it. I did multi-sample alignment for every segment separately to consider resortment, and then concatenated the aligned sequences horizontally.

[roblanf]

If I've understood that correctly, that will still cause huge issues if there's reassortment. The analyses you're doing assume there's a single tree (and set of dates) underlying the entire alignment. You'd want to do some topology tests first at a minimum (I'm guessing there are probably better ways of testing for reassortment in the flu literature though!).

[tahirsajid]

Thank, Rob for making it clear for me. I am new to Phylogenetic analyses of viruses - learning new things. If I understand your comment correctly, you mean that I should check per segment topologies first. If consistent across all segments, then I should be confident on whole genome tree. Also, if you can advise, does this scenario apply to time tree and substitutation rate tree both?
Thanking for your advice.

[roblanf]

Yes, exactly. You can only concatenate if you are willing to assume the trees are identical. This is particularly critical for flu viruses, where the segments can reassort freely. I assume that the flu literature (which I don't know) has sensible tests for this, but if you can't find anything there's a lovely principled way to do it in this paper:

https://academic.oup.com/sysbio/article-abstract/57/1/104/1696657

[tahirsajid]

Thank you, Rob, It will help a lot.
Cheers.