PTF-Vāc:Ab-initiodiscovery of plant transcription factors binding sites using explainable and generative deep co-learning encoders-decoders

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 31, 2024

Abstract Discovery of transcription factors (TFs) binding sites (TFBS) and their motifs in plants pose significant challenges due to high cross-species variability. The interaction between TFs is highly specific context dependent. Most the existing TFBS finding tools are not accurate enough discover these plants. They fail capture variability, interdependence TF structure its TFBS, specificity binding. Since they coupled predefined model/matrix, vulnerable towards volume quality data provided build motifs. All software make a presumption that user input would be any particular which renders them very limited uses. This all makes hardly use for purposes like genomic annotations newly sequenced species. Here, we report an explainable Deep Encoders-Decoders generative system, PTF-Vāc, founded on universal model deep co-learning variability structure, PTFSpot, making it completely free from bottlenecks mentioned above. It has successfully decoupled process discovery prior step motif requirement models. Due TF:DNA interactions as guide, can total independence volume, species PTF-Vāc accurately detect even never seen before families species, used define credible report.

Language: Английский

PTFSpot: Deep co-learning on transcription factors and their binding regions attains impeccable universality in plants

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Nov. 20, 2023

Abstract Unlike animals, variability in transcription factors (TF) and their binding regions (TFBR) across the plants species is a major problem which most of existing TFBR finding software fail to tackle, rendering them hardly any use. This limitation has resulted into underdevelopment plant regulatory research rampant use Arabidopsis like model species, generating misleading results. Here we report revolutionary transformers based deep-learning approach, PTFSpot, learns from TF structures co-variability bring universal TF-DNA interaction detect with complete freedom specific models’ limitations. During series extensive benchmarking studies over multiple experimentally validated data, it not only outperformed by >30% lead, but also delivered consistently >90% accuracy even for those families were never encountered during building process. PTFSpot makes possible now accurately annotate TFBRs genome total lack information, completely free bottlenecks models.

Language: Английский