Ordering taxa in image convolution networks improves microbiome-based machine learning accuracy

Shtossel, Oshrit; Isakov, Haim; Turjeman, Sondra; Koren, Omry; Louzoun, Yoram

doi:10.1080/19490976.2023.2224474

Cited by 5 publications

(4 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As the first step, miMic preprocesses the microbiome frequencies using MIPMLP [ 41 ] (see Methods ) and translates them into a cladogram of means using the iMic algorithm [ 42 ] (Fig. 1 D data processing).…”

Section: Resultsmentioning

confidence: 99%

“…The host microbiome has been associated with a myriad of phenotypes. This association is often performed using three main arguments: A) Samples with a given condition are closer one to each other than to samples without this label [ 48 , 49 ], B) the microbiome samples can be used to predict the phenotye using machine learning methods [ 42 , 50 , 51 ], or C) specific taxa are associated with a condition/property/label of the host [ 11 , 12 , 14 , 17 ]. This last approach has been termed differential abundance analysis, and a large number of methods have been proposed to perform it.…”

Section: Discussionmentioning

confidence: 99%

“…More specifically, miMic first applies normalization and translation of ASV to log-normalized taxa frequencies using the MIPMLP pipeline [ 41 ]. These taxa are combined to a cladogram of means[ 42 ] (see detailed explanation in the “ Methods ” and Fig. 1 D Data processing step).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

mi-Mic: a novel multi-layer statistical test for microbiota-disease associations

Shtossel,

Finkelstein,

Louzoun

2024

Genome Biol

Self Cite

View full text Add to dashboard Cite

Abstractmi-Mic, a novel approach for microbiome differential abundance analysis, tackles the key challenges of such statistical tests: a large number of tests, sparsity, varying abundance scales, and taxonomic relationships. mi-Mic first converts microbial counts to a cladogram of means. It then applies a priori tests on the upper levels of the cladogram to detect overall relationships. Finally, it performs a Mann-Whitney test on paths that are consistently significant along the cladogram or on the leaves. mi-Mic has much higher true to false positives ratios than existing tests, as measured by a new real-to-shuffle positive score.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

mi-Mic: a novel multi-layer statistical test for microbiota-disease associations

Shtossel,

Finkelstein,

Louzoun

2024

Genome Biol

Self Cite

View full text Add to dashboard Cite

show abstract

“…These models’ predictions were more accurate than the simple models’ predictions. The highest SCC was obtained by applying the iMic model [ 62 ] to the donors’ MIPMLP preprocessed taxa frequencies, with an SCC of 0.6 +/− 0.004 and an R2 score of 0.358 +/− 0.003 on the Shannon diversity (Fig. 2 A, B pink bars).…”

Section: Resultsmentioning

confidence: 99%

Recipient-independent, high-accuracy FMT-response prediction and optimization in mice and humans

Shtossel,

Turjeman,

Riumin

et al. 2023

Microbiome

Self Cite

View full text Add to dashboard Cite

Background Some microbiota compositions are associated with negative outcomes, including among others, obesity, allergies, and the failure to respond to treatment. Microbiota manipulation or supplementation can restore a community associated with a healthy condition. Such interventions are typically probiotics or fecal microbiota transplantation (FMT). FMT donor selection is currently based on donor phenotype, rather than the anticipated microbiota composition in the recipient and associated health benefits. However, the donor and post-transplant recipient conditions differ drastically. We here propose an algorithm to identify ideal donors and predict the expected outcome of FMT based on donor microbiome alone. We also demonstrate how to optimize FMT for different required outcomes. Results We show, using multiple microbiome properties, that donor and post-transplant recipient microbiota differ widely and propose a tool to predict the recipient post-transplant condition (engraftment success and clinical outcome), using only the donors’ microbiome and, when available, demographics for transplantations from humans to either mice or other humans (with or without antibiotic pre-treatment). We validated the predictor using a de novo FMT experiment highlighting the possibility of choosing transplants that optimize an array of required goals. We then extend the method to characterize a best-planned transplant (bacterial cocktail) by combining the predictor and a generative genetic algorithm (GA). We further show that a limited number of taxa is enough for an FMT to produce a desired microbiome or phenotype. Conclusions Off-the-shelf FMT requires recipient-independent optimized FMT selection. Such a transplant can be from an optimal donor or from a cultured set of microbes. We have here shown the feasibility of both types of manipulations in mouse and human recipients.

show abstract

Deciphering the gut microbiome: The revolution of artificial intelligence in microbiota analysis and intervention

Abavisani,

Khoshrou,

Foroushan

et al. 2024

Current Research in Biotechnology

View full text Add to dashboard Cite

Ordering taxa in image convolution networks improves microbiome-based machine learning accuracy

Cited by 5 publications

References 66 publications

mi-Mic: a novel multi-layer statistical test for microbiota-disease associations

mi-Mic: a novel multi-layer statistical test for microbiota-disease associations

Recipient-independent, high-accuracy FMT-response prediction and optimization in mice and humans

Deciphering the gut microbiome: The revolution of artificial intelligence in microbiota analysis and intervention

Contact Info

Product

Resources

About