Preprint Identification of a multi-omics factor predictive of long COVID in the IMPACC study, 2025, Gabernet et al.

SNT Gatchaman · Feb 15, 2025

Identification of a multi-omics factor predictive of long COVID in the IMPACC study
Gisela Gabernet; Jessica Maciuch; Jeremy P Gygi; John F Moore; Annmarie Hoch; Caitlin Syphurs; Tianyi Chu; Naresh Doni Jayavelu; David B Corry; Farrah Kheradmand; Lindsey R Baden; Rafick-Pierre Sekaly; Grace A McComsey; Elias K Haddad; Charles B Cairns; Nadine Rouphael; Ana Fernandez-Sesma; Viviana Simon; Jordan P Metcalf; Nelson I Agudelo Higuita; Catherine L Hough; William B Messer; Mark M Davis; Kari C Nadeau; Bali Pulendran; Monica Kraft; Chris Bime; Elaine F Reed; Joanna Schaenman; David J Erle; Carolyn S Calfee; Mark A Atkinson; Scott C Brackenridge; Esther Melamed; Albert C Shaw; David A Hafler; Al Ozonoff; Steven E Bosinger; Walter Eckalbar; Holden T Maecker; Seunghee Kim-Schulze; Hanno Steen; Florian Krammer; Kerstin Westendorf; IMPACC Network; Bjoern Peters; Slim Fourati; Matthew C Altman; Ofer Levy; Kinga K Smolen; Ruth R Montgomery; Joann Diray-Arce; Steven H Kleinstein; Leying Guan; Lauren I R Ehrlich

Following SARS-CoV-2 infection, ~10-35% of COVID-19 patients experience long COVID (LC), in which often debilitating symptoms persist for at least three months. Elucidating the biologic underpinnings of LC could identify therapeutic opportunities.

We utilized machine learning methods on biologic analytes and patient reported outcome surveys provided over 12 months after hospital discharge from >500 hospitalized COVID-19 patients in the IMPACC cohort to identify a multi-omics "recovery factor".

IMPACC participants who experienced LC had lower recovery factor scores compared to participants without LC. Biologic characterization revealed increased levels of plasma proteins associated with inflammation, elevated transcriptional signatures of heme metabolism, and decreased androgenic steroids in LC patients. The recovery factor was also associated with altered circulating immune cell frequencies.

Notably, recovery factor scores were predictive of LC occurrence in patients as early as hospital admission, irrespective of acute disease severity. Thus, the recovery factor identifies patients at risk of LC early after SARS-CoV-2 infection and reveals LC biomarkers and potential treatment targets.

Link | PDF (Preprint: BioRxiv) [Open Access]

Hutan · Feb 15, 2025

I think this study may suffer from an overly permissive view of what Long Covid is. That, coupled with their hospitalised samples means that the biochemical markers are not necessarily relevant to LC ME/CFS.

The Recovery factor is the combination of the parameters that best differentiated their model training recovered versus LC cohorts. Higher 'Recovery Factor' values are better. I don't think the results are overly impressive.

They only had about 500 people in their sample, and only 80 of those seem to have the LC 'physical deficit' that they built their model on. Slice and dice that into the 80% training and 20% test, and take into account that they had a very large number of measurements (nearly 7000) and I don't think they got very good separation of the LC and Recovered groups.

We focused on predicting LC in the Convalescent cohort from the multi-omics immune profiling data collected during the convalescent phase. Since the binary LC labels (presence or absence of LC) per participant could omit valuable information captured by the numeric PRO measures at each participant visit, we constructed separate SPEAR models to generate supervised factors including PRO measure scores (SPEAR Physical, SPEAR Cognitive, SPEAR Mental, SPEAR Impact, SPEAR Dyspnea) or the LC binary labels (SPEAR LC) as response variables (Figure S2A).
Click to expand...

The lasso model trained on the SPEAR Physical multi-omics factor achieved the highest predictive performance as evaluated with the area under the receiver- operating characteristic curve (AUROC) (Figure 2A).
Click to expand...

As far as I can see, they subsetted their LC cohort, and then played around with models, selecting the model that performed best for them. So, for example, the model predicting people who had ongoing breathlessness didn't make the cut. It's just another opportunity for bias.

Hutan · Feb 15, 2025

The SPEAR Physical model identified 26 analytes across four assays that were significant in the recovery factor (SPEAR Bayesian posterior selection probability ≥ 0.95), and we performed individual associations of these features with LC status in the test cohort, adjusting for age and sex (Figure 3B, Figure S5).
Click to expand...

Of these, DNER (Delta And Notch-Like Epidermal Growth Factor- Related Receptor), a non-canonical Notch ligand that has been implicated in promoting tumor growth and metastasis and in supporting wound healing48,49 was significantly reduced in LC participants, consistent with a prior study of plasma proteomics in LC subjects28. The remaining serum Olink analytes were negatively associated with the recovery factor.

In particular, they included proteins and cytokines associated with chronic inflammatory conditions50–55, particularly endothelial/vascular inflammation (FGF23, FGF21, CXCL9, TNFRSF11B and TNFRSF9 (CD137)), as well as inflammation-associated myeloid regulators56–58 (MMP10 and CSF1). Elevated levels of IL10RB have been previously associated with worse outcomes in acute COVID- 19 infection59, consistent with elevation under inflammatory conditions. LRG1, a protein elevated in LC participants, is induced by IL-6 and other inflammatory cytokines and has been implicated in angiopathic activity60–62.

Phenylacetylglutamate and phenylacetylglutamine are gut microbiota- derived metabolites associated with vascular inflammation and thrombosis63. Finally, the OSBP2 (ORP4) transcript, which encodes an oxysterol binding protein64, was a leading edge gene in the Hallmark Heme Metabolism gene set that was elevated in LC participants.
Click to expand...

Hutan · Feb 15, 2025

androgenic steroids

Several metabolites from the androgenic steroids pathway were represented in the 26 significant analytes and were positively associated with the recovery factor, indicating higher levels correlate with better physical function. When we tested these metabolites for their individual association with LC status, five (DHEA-S, epiandrosterone sulfate, androsterone sulfate, 5alpha-androstan- 3beta,17beta-diol monosulfate (2), 5alpha-androstan-3beta,17alpha-diol disulfate) were significantly lower in LC participants, adjusting for age and sex (Figure 3B).

Androgens can suppress inflammation65, suggesting that the higher level of androgenic steroids in participants of the MIN group could reflect better control of chronic inflammation. These findings are consistent with prior reports showing lower levels of sex hormones in LC31. Five metabolites related to pregnenolone were also represented in the significant SPEAR analytes (Figure 3B).

Pregnenolone is synthesized from cholesterol as the first step of the steroid hormone biosynthesis pathway and is known to have potent effects as an inhibitor of inflammation66 and as a neurosteroid67. Altogether, these findings are consistent with a prominent role for persistent inflammation in LC with dysregulation of key analytes that may contribute to symptoms in LC, including elements that drive angiopathy, reduce wound healing, and alter heme metabolism.
Click to expand...

Figure 3b shows that of the 15 or so steroids, cortisol is not one found to be different.

________
73 unique features they believe are particularly useful in separating the recovered and LC groups

The feature sets from heme metabolism and androgenic steroids identified by GSEA analysis combined with the significant SPEAR analytes represent 73 unique features that potentially condense the predictive power of the recovery factor into a smaller feature set. To test this hypothesis, we calculated the geometric mean of the 43 leading edge heme metabolism and 12 androgenic steroid features, as well as the 26 significant SPEAR analytes. All three geometric mean scores were independently significantly associated with LC in the test cohort (Figure 3C). Furthermore, the combined score that includes analytes from all three feature sets discriminates MIN and LC participants with even greater significance (Figure 3C). Thus, while the recovery factor is comprised of weighted contributions from 6,807 features, we have identified a smaller set of 73 unique features that discriminates participants according to LC status in the convalescent period.
Click to expand...

Hutan · Feb 15, 2025

Consistent with our finding, the Hallmark Heme Metabolism pathway was previously reported by Hanson et al.29 as an enriched pathway in participants with persisting symptoms 1-3 months after acute SARS-CoV-2 infection compared to participants without persisting symptoms. This cohort comprised 102 participants, including non-hospitalized and hospitalized individuals29. To determine whether the same heme metabolism-related genes were dysregulated in LC participants in the IMPACC and Hanson et al. cohorts, we used the leading edge genes from the significant Hallmark Heme Metabolism pathway in our GSEA analysis (Figure S5A) and calculated the geometric mean gene expression in PBMCs from the Hanson et al. cohort29. We found that our heme metabolism leading edge genes significantly differentiated participants with persistent vs. resolved symptoms after COVID-19 infection at multiple time points in the independent cohort (Figure S5C), validating the reproducibility of the gene expression datasets and underscoring the importance of this subset of heme metabolism genes.
Click to expand...

mariovitali · Feb 23, 2025

With the help of Grok 3, I realized that the study found several metabolites below that are related to dehydrotestosterone and more specifically 5 alpha reductase - related metabolites. Given my story ( I got ME/ CFS from Finasteride which is a 5 alpha reductase inhibitor) these results may be yet another important piece of the puzzle . Here is a part of the answer from Grok3

The presence of 5α-androstan-3β,17β-diol monosulfate and 5α-androstan-3β,17α-diol disulfate indicates involvement of 5α-reductase activity (which converts testosterone or DHEA intermediates into 5α-reduced compounds) followed by sulfation
Click to expand...

Log in or Sign up

Preprint Identification of a multi-omics factor predictive of long COVID in the IMPACC study, 2025, Gabernet et al.

SNT Gatchaman Senior Member (Voting Rights)

Hutan Moderator Staff Member

Hutan Moderator Staff Member

Hutan Moderator Staff Member

Hutan Moderator Staff Member

mariovitali Senior Member (Voting Rights)

Share This Page

Log in or Sign up

Preprint Identification of a multi-omics factor predictive of long COVID in the IMPACC study, 2025, Gabernet et al.

SNT Gatchaman Senior Member (Voting Rights)

Hutan Moderator Staff Member

Hutan Moderator Staff Member

Hutan Moderator Staff Member

Hutan Moderator Staff Member

mariovitali Senior Member (Voting Rights)

Share This Page

Useful Searches