MFMDScen in the MMGFM Framework

• MFMDScen is an informal shorthand linked to MMGFM, highlighting simulation scenarios rather than defining a distinct method.
• The MMGFM model integrates multiple studies and modalities by separating shared and study-specific latent factors with covariate adjustments.
• Simulation scenarios demonstrate MMGFM’s robustness and superior performance in handling complex, mixed-type data compared to alternative methods.

Within the context of (Liu et al., 14 Jul 2025), MFMDScen is not a defined acronym. The source paper explicitly introduces MMGFM, a high-dimensional multi-study, multi-modality, covariate-augmented generalized factor model, and uses “scenario” only for the simulation section’s Scenarios 1–4. Accordingly, the technically grounded interpretation of MFMDScen is as an apparent mislabeling or informal shorthand connected to the MMGFM framework rather than a distinct method. MMGFM is designed for settings with multiple studies, multiple modalities, mixed variable types, and additional covariates, and it combines study-shared, study-specific, and modality-related latent structure in a single generalized factor-analysis model (Liu et al., 14 Jul 2025).

1. Terminological status and scope

The cited work does not define the term MFMDScen. Instead, it defines MMGFM and frames its contribution as addressing a gap left by methods that focus predominantly either on multi-study integration or on multi-modality integration, but not on data with diverse modalities measured across multiple studies simultaneously (Liu et al., 14 Jul 2025).

This distinction matters for interpretation. If MFMDScen is encountered in connection with this paper, the evidence supports two constrained readings. First, it may refer indirectly to MMGFM itself. Second, it may refer to the paper’s simulation Scenarios 1–4, since “scenario” language appears there but not as a named acronym. A plausible implication is that MFMDScen should not be treated as an established methodological term unless further source material defines it explicitly.

2. Data structure and model specification

MMGFM is formulated for data from multiple studies $s=1,\dots,S$, each containing multiple modalities $m=1,\dots,M$, with potentially different variable types across modalities. For study $s$ and modality $m$, the observed matrix is

$$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$

with covariates $z_{si}\in\mathbb{R}^d$. The hierarchical generalized factor-analysis form is

$$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$

The latent decomposition is organized as follows:

• $f_i$: study-shared latent factor
• $\lambda_{mj}$: corresponding loading
• $g_{si}$: study-specific latent factor
• $m=1,\dots,M$0: corresponding study-specific loading
• $m=1,\dots,M$1: study-specific, modality-shared scalar factor
• $m=1,\dots,M$2: overdispersion/random effect term

The latent variables are Gaussian:

$m=1,\dots,M$3

and

$m=1,\dots,M$4

with $m=1,\dots,M$5 diagonal. The overdispersion term is included to capture within-modality correlation and extra variation for non-Gaussian outcomes (Liu et al., 14 Jul 2025).

For the most emphasized modalities, the model treats three variable types:

• continuous: $m=1,\dots,M$6
• count: $m=1,\dots,M$7
• binary/categorical: $m=1,\dots,M$8 is binomial/logistic-style with

$m=1,\dots,M$9

This jointly accommodates multi-study structure, multi-modality structure, mixed variable types, covariate effects, and study-shared, study-specific, and modality-specific variation.

3. Latent decomposition, covariates, and identifiability

A central feature of MMGFM is the separation of latent signal into interpretable components. Study-shared factors $s$0 represent latent structure common across all studies; their loadings $s$1 are modality- and variable-specific but not study-specific. Study-specific factors $s$2 represent latent variation unique to study $s$3, with loadings $s$4 varying by study and modality. The modality-shared random effect $s$5 captures correlation among variables within the same modality and study. Covariates enter through the linear term $s$6, allowing measured information to explain part of the signal rather than being absorbed into latent factors (Liu et al., 14 Jul 2025).

The paper explicitly states that the model is not identifiable without constraints, and that identifiability conditions are established in Appendix A; it also notes that Condition (C1) in the asymptotic section guarantees identifiability. This is significant because factor models otherwise admit arbitrary rotations, sign changes, and confounding between covariate effects and latent factors. In this setting, identifiability underwrites the interpretation of what is shared across studies, what is study-specific, and how covariates contribute.

In the simulation section, identifiability is also enforced in data generation through SVD-based normalization of loading matrices, for example

$s$7

followed by block selection to form shared and specific loadings. This suggests that interpretability is not treated as a purely formal property but as a design constraint affecting both theory and empirical evaluation.

4. Variational approximation and estimation procedure

The observed log-likelihood for observation $s$8 in study $s$9 is

$m$0

where $m$1 collects the model parameters in study $m$2. The source paper characterizes this likelihood as analytically intractable because it integrates over four large latent random matrices/vectors (Liu et al., 14 Jul 2025).

To address this, it introduces a mean-field variational approximation

$m$3

with a fully factorized Gaussian variational family. The corresponding variational lower bound is

$m$4

By Jensen’s inequality,

$m$5

with equality if and only if the variational density equals the true posterior. The variational posterior therefore acts as the best approximation to the intractable posterior within the chosen mean-field family.

Estimation proceeds via a variational EM algorithm. In the E-step, the variational parameters are updated by

$m$6

Because $m$7 are not conjugate for the non-Gaussian likelihood, the update uses Laplace approximation combined with Taylor approximation. In the M-step, the profiled variational objective

$m$8

is maximized over model parameters. The paper defines the maximum variational lower bound estimator and the maximum variational log-likelihood estimator, and proves that they are equal. It also states that for each $m$9, $$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$0 is the unique maximizer of the lower bound.

5. Asymptotic properties and factor-number selection

The asymptotic theory profiles out the variational parameters and treats the resulting objective as an M-estimation problem. Under conditions $$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$1–$$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$2, with $$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$3 and $$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$4, Theorem 3 gives the rates

$$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$5

and

$$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$6

The paper interprets these rates as reflecting information pooling: shared/modality-wide parameters are estimated using all $$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$7 samples, whereas study-specific parameters are informed only by $$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$8 observations from study $$X_{sm}=(x_{simj})_{i\le n_s,\; j\le p_m}\in\mathbb{R}^{n_s\times p_m},$$9 (Liu et al., 14 Jul 2025).

Theorem 4 provides an asymptotic linear expansion,

$z_{si}\in\mathbb{R}^d$0

and asymptotic normality for parameter blocks. Examples given in the source include

$z_{si}\in\mathbb{R}^d$1

with analogous results for $z_{si}\in\mathbb{R}^d$2, and for study- and modality-specific parameters,

$z_{si}\in\mathbb{R}^d$3

$z_{si}\in\mathbb{R}^d$4

$z_{si}\in\mathbb{R}^d$5

For dimensionality selection, the paper proposes a step-wise singular value ratio (SVR) criterion. After fitting with generous upper bounds $z_{si}\in\mathbb{R}^d$6 and $z_{si}\in\mathbb{R}^d$7, and writing $z_{si}\in\mathbb{R}^d$8 for the estimated loading matrix for modality $z_{si}\in\mathbb{R}^d$9, the shared factor number is estimated by

$$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$0

where $$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$1 is the $$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$2-th largest singular value. If the $$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$3’s differ across modalities, the estimator is

$$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$4

otherwise, $$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$5 is taken as the mode of $$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$6. With the selected $$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$7, the model is refit and the same SVR principle is applied to study-specific loading matrices $$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$8 to obtain $$x_{simj}\mid y_{simj} \stackrel{c.i.d.}{\sim} EF(g_m(y_{simj})),\qquad y_{simj} = \tau_{sim} + z_{si}^\top \beta_{mj} + \lambda_{mj}^\top f_i + \gamma_{smj}^\top g_{si} + v_{sim} + \varepsilon_{simj}.$$9.

6. Simulation evidence, real-data application, and software

The simulation program is organized into Scenarios 1–4, which likely explains why an informal label such as MFMDScen might arise, although the paper itself does not define that term. In Scenario 1, MMGFM is compared with GFM, MRRR, MSFR, and MultiCOAP in a three-study, three-Poisson-modality setting with covariates and varying $f_i$0. The reported findings are that MMGFM consistently outperforms competitors in estimating factors and regression coefficients, remains robust as intramodality correlation $f_i$1 increases, and can still estimate loadings well even when within-modality correlation is present, whereas methods that ignore this structure degrade substantially (Liu et al., 14 Jul 2025).

In Scenario 2, which uses mixed Gaussian and Poisson modalities, MMGFM achieves the best mean trace statistics and smallest coefficient error. Scenario 3 varies modality types, sample size, dimension, number of studies, and number of modalities; the summary reported in the source is that MMGFM remains dominant across these settings, while competing methods often break down or perform poorly as the problem becomes more complex. Scenario 4 evaluates factor-number selection and reports that the step-wise SVR method accurately identifies $f_i$2 and $f_i$3, with accuracy improving as sample size increases and declining as noise or dimension increases.

The real-data application analyzes CITE-seq single-cell multimodal sequencing data from 12 PBMC subjects with COVID-19 status: 4 severe, 3 moderate, and 5 healthy. These three groups are treated as three studies. The two modalities are gene expression counts and CLR-normalized protein markers, and the covariates include age, sex, and days since symptom onset. Using the proposed factor-selection criterion, the fitted dimensions are

$f_i$4

The reported outcomes are that MMGFM captures both gene and protein information well, outperforms methods that only handle one modality type or cannot separate study-specific structure, and produces extracted features that support joint clustering and biological interpretation. The study-specific loading matrices are further used to identify potentially important genes and proteins related to immune response and COVID severity; the inferred clusters are said to align with known biology.

The method is implemented in the publicly available R package MMGFM on CRAN. The package provides functionality for fitting the model, performing variational EM estimation, and carrying out factor-number selection via the step-wise SVR rule.