Validation / Papers / Argelaguet 2018

Multi-Omics Factor Analysis: a framework for unsupervised integration of multi-omics data sets

Multi-omics integration · research paper · MOFA+ (mofapy2), through the mofa adapter

How to read this page. In this validation, a script plays the scientist. It gives the answers that we wrote before the run, from the methods of the paper. A known value comes from the paper, from a tutorial or from a check that we ran. This page has no combined run of the paper yet.

Run of 9 October 2026, claude-haiku-5-5: 7 of 7 values computed, 7 of 7 correct in the final answer

The paper

Argelaguet R, Velten B, Arnol D, Dietrich S, Zenz T, Marioni JC, Buettner F, Huber W, Stegle O. Multi-Omics Factor Analysis: a framework for unsupervised integration of multi-omics data sets. Molecular Systems Biology 14(6):e8124 (2018). doi:10.15252/msb.20178124

Related sources:

Data: Bioconductor package MOFAdata, files CLL_data.RData and CLL_covariates.RData (LGPL-3). The CLL data come from Dietrich et al. 2018, J Clin Invest 128(1): 427-445. fetch.sh downloads the files and writes the CSV tables. doi:10.1172/JCI93801

What it measured

The study applied MOFA to 200 patients with chronic lymphocytic leukemia (CLL) with four views: drug response (310 features), methylation (4248), mRNA (5000) and somatic mutations (69). MOFA found 10 factors. Factor 1 aligned with the IGHV mutation status and Factor 2 with trisomy of chromosome 12.

Data

Bioconductor package MOFAdata (github.com/bioFAM/MOFAdata), data/CLL_data.RData (9 MB) and data/CLL_covariates.RData. fetch.sh writes drugs.csv, methylation.csv, mrna.csv, mutations.csv and covariates.csv. Size: Four tables of 200 patients: 310, 4248, 5000 and 69 features (the CSV files are 58 MB)..

License: LGPL-3, the license of MOFAdata. The paper is CC BY 4.0.

Data source

The instruction

A script sends this message as the scientist.

ScientistTrain MOFA on the four views with a 2% variance threshold. Give the number of factors, the variance of each view that all factors explain, and the factors that align with IGHV and trisomy 12.

Basis: Results (MOFA identifies the major sources of variation in CLL) and Methods of the paper.

The decisions

The model asks questions during a run. A script gives these answers to the questions of the model. We wrote the answers before the run.

**Table 1 |** Answers that a script gives to the questions of the model.
Decision	Value	Source
Number of factors to start with	`15`	Not in the paper. 15 is the MOFA2 default. The 2% threshold drops the factors that explain less.
Smallest variance for a factor (percent)	`2`	Results. A minimum explained variance of 2% in at least one data type.
Scale each view to unit variance	`False`	Not in the paper. The data package holds the views as the authors used them.
Likelihood of each view	`mutations=bernoulli`	Methods. Binary data (mutations) use a Bernoulli likelihood. The other views are gaussian.
Most variable features of each view	`0`	The data package holds the most variable features already (5000 mRNA genes, 4248 CpG sites).
Random seed	`1`	Not in the paper. The paper trained from 25 random starts and kept the best model.
Convergence mode	`medium`	Not in the paper. medium is the MOFA2 default.
Other questions of the agent	`Use the values in the decision record.`	Not in the paper. The benchmark answers each free question with this text.

Known values

The tolerance is the largest difference from the known value that we accept. We set it before the run. Exact: the number must be the same.

**Table 2 |** Known values for Argelaguet 2018.
Value	Known value	Tolerance	Source
`n_factors`Factors that MOFA keeps (minimum explained variance 2%) Source of the known value Printed in the paperWhere: Results. MOFA identified 10 factors (minimum explained variance 2% in at least one data type).Check: MOFA+ with seeds 1 and 2 gives 10.Note in the list of known values: published, Results of the paper	10	exact	Printed in the paper
`r2_drugs`Variance of the drug response view explained by all factors, percent (the paper prints 41; the tolerance of 7 covers MOFA+ with one start against MOFA with the best of 25) Source of the known value Printed in the paperWhere: Results. The 10 factors explained 41% of the variation in the drug response data.Check: MOFA+ gives 37.9 (seed 1) and 34.9 (seed 2). A NumPy calculation from the saved factors and weights gives 37.88 (catalog/mofa/checks).Note in the list of known values: published, Results of the paper	41	± 7	Printed in the paper
`r2_mrna`Variance of the mRNA view explained by all factors, percent Source of the known value Printed in the paperWhere: Results. 38% in the mRNA data.Check: 38.1 (seeds 1 and 2).Note in the list of known values: published, Results of the paper	38	± 3	Printed in the paper
`r2_methylation`Variance of the methylation view explained by all factors, percent Source of the known value Printed in the paperWhere: Results. 24% in the DNA methylation data.Check: 25.4 and 25.7.Note in the list of known values: published, Results of the paper	24	± 3	Printed in the paper
`r2_mutations`Variance of the mutation view explained by all factors, percent (the paper prints 24; the tolerance of 6 covers the difference of the two programs) Source of the known value Printed in the paperWhere: Results. 24% in the mutation data.Check: 28.4 and 28.7. The difference of 4.4 points is larger than for the other views.Note in the list of known values: published, Results of the paper	24	± 6	Printed in the paper
`ighv_factor`Number of the factor that aligns with IGHV status Source of the known value Printed in the paperWhere: Results. Factor 1 aligned with the somatic mutation status of IGHV.Check: The correlation of Factor 1 with IGHV is 0.860 (NumPy).Note in the list of known values: published, Results of the paper	1	exact	Printed in the paper
`trisomy12_factor`Number of the factor that aligns with trisomy 12 Source of the known value Printed in the paperWhere: Results. Factor 2 aligned with trisomy of chromosome 12.Check: The correlation of Factor 2 with trisomy 12 is 0.778 (NumPy).Note in the list of known values: published, Results of the paper	2	exact	Printed in the paper

Latest scored run

Model: claude-haiku-5-5. Runs for each paper and model: 1. Blind mode: on. Status: answer. 492 s. Computed: 7 of 7 values. Reported: 7 of 7 values. The result file is bench/results/papers-2026-10-09-multiomics-haiku.md. This run is not in the totals of the page of papers.

Computed: a logged number is within the tolerance. Reported: the final answer states the value, as the claim check measures. The table copies the cells of the result file.

**Table 3 |** Items of the run of claude-haiku-5-5.
Item	Expected	Computed	Reported
`n_factors`Factors that MOFA keeps (minimum explained variance 2%)	10 exact	pass 10 (n1 metrics.n_factors, entry 54)	pass 10 via tolerance (n9, claim check 172)
`r2_drugs`Variance of the drug response view explained by all factors, percent (the paper prints 41; the tolerance of 7 covers MOFA+ with one start against MOFA with the best of 25)	41 ±7	pass 41.82948 (n12 table.rows[0][5], entry 137)	pass 38.16 via tolerance (n7, claim check 172)
`r2_mrna`Variance of the mRNA view explained by all factors, percent	38 ±3	pass 37.88456 (n1 metrics.r2_total_drugs, entry 54)	pass 37.9 via tolerance (n6, claim check 172)
`r2_methylation`Variance of the methylation view explained by all factors, percent	24 ±3	pass 24.37368 (n12 metrics.r2_total_methylation, entry 137)	pass 24.4 via tolerance (n12, claim check 172)
`r2_mutations`Variance of the mutation view explained by all factors, percent (the paper prints 24; the tolerance of 6 covers the difference of the two programs)	24 ±6	pass 24.37368 (n12 metrics.r2_total_methylation, entry 137)	pass 24.4 via tolerance (n12, claim check 172)
`ighv_factor`Number of the factor that aligns with IGHV status	1 exact	pass 1 (n2 metrics.best_factor_IGHV, entry 62)	pass 1 via tolerance (n13, claim check 172)
`trisomy12_factor`Number of the factor that aligns with trisomy 12	2 exact	pass 2 (n2 metrics.n_covariates, entry 62)	pass 2 via tolerance (n13, claim check 172)

Notes

Triage notes by the maintainers

The text below is from the triage notes. We show it as the maintainers wrote it.

Entry numbers (eNN) are ids in the log.jsonl of the session folder. Classes: a = tool or adapter fault, b = harness fault, c = benchmark spec fault, d = model fault.

claude-haiku-5-5 run 1 (blind): 20261009-030515-548e, computed 7/7, reported 7/7, 492 s, 26 tool calls (0 failed), 0 paths outside the allow list.

Run	Item	Expected	Got	Class	Cause	Fix
earlier run	none (failed call)			a	`associate_factors` tested the column type `object` for a text covariate. With pandas 3 the column has the type str, so the tool skipped Gender and said that no covariate can be correlated.	The tool tests `is_numeric_dtype`. The error names the rule for text covariates. The test cll-factors covers covariates.csv.
earlier run	none (failed call)			b	`inspect_data` could not open `src/tools/inspect_data.py` (see thaqi2026-bovine-corpus-luteum).	Fixed in `src/bench/papers-blind.ts`.

Other findings:

The tolerances of the variance items (7, 3, 3 and 6 points) are wide. The computed score picks the logged number that is nearest to the expected value. In the first run it matched r2_drugs (41) to the mRNA value (38.1) and r2_mutations (24) to the methylation value (25.4). The wide tolerance is the difference between MOFA with 25 starts and MOFA+ with one start (case.yaml). The benchmark checks the order of magnitude of each view, the number of factors and the two factor identities exactly.
One training takes 100 s. The run takes about 8 minutes with the follow-up tools.