ProNA2020 predicts protein–DNA, protein–RNA, and protein–protein binding proteins and residues from sequence

Qiu, Jiajun; Bernhofer, Michael; Heinzinger, Michael; Kemper, Sofie; Norambuena, Tomás; Melo, Francisco S.; Rost, Burkhard

doi:10.1016/j.jmb.2020.02.026

Cited by 83 publications

(102 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(2) De-novo methods predict protein function through machine learning 8 . If applicable, the first approach tends to out-perform the second 16-19 although it largely misses discoveries 20 . The progress of computational methods has been monitored by CAFA ( Critical Assessment of protein Function Annotation algorithms ) 12,21,22 , an international collaboration to advancing and assessing methods that bridge the sequence-annotation gap.…”

Section: Introductionmentioning

confidence: 99%

Embeddings from deep learning transfer GO annotations beyond homology

Littmann

Heinzinger

Dallago

et al. 2020

Preprint

Self Cite

109

View full text Add to dashboard Cite

Knowing protein function is crucial to advance molecular and medical biology, yet experimental function annotations through the Gene Ontology (GO) exist for fewer than 0.5% of all known proteins. Computational methods bridge this sequence-annotation gap typically through homology-based annotation transfer by identifying sequence-similar proteins with known function or through prediction methods using evolutionary information. Here, we proposed predicting GO terms through annotation transfer based on proximity of proteins in the SeqVec embedding rather than in sequence space. These embeddings originated from deep learned language models (LMs) for protein sequences (SeqVec) transferring the knowledge gained from predicting the next amino acid in 250 million protein sequences. Replicating the conditions of CAFA3, our method reached an Fmax of 37± 2%, 50± 3%, and 57± 2% for BPO, MFO, and CCO, respectively. This was numerically close to the top ten methods that had participated in CAFA3. Restricting the annotation transfer to proteins with <20% pairwise sequence identity to the query, performance dropped (Fmax BPO 33± 2%, MFO 43± 3%, CCO 53± 2%); this still outperformed naïve sequence-based transfer. Preliminary results from CAFA4 appeared to confirm these findings. Overall, this new method may help, in particular, to annotate novel proteins from smaller families or proteins with intrinsically disordered regions.

show abstract

Section: Introductionmentioning

confidence: 99%

Embeddings from deep learning transfer GO annotations beyond homology

Littmann

Heinzinger

Dallago

et al. 2020

Preprint

Self Cite

109

View full text Add to dashboard Cite

show abstract

“…The total over both types of experimental annotations (binding/effect) provided the upper limit for SAVs with an experimental annotation about either binding or effect or both, namely 79,397 SAVs (1.2%): 404 of these for common SAVs and 78,993 for rare SAVs (2nd to last row labelled SUM experimental) interfaces with common SAVs (18 of 34,309, i.e. 0.05%).Therefore, results had to be based on a prediction method, namely ProNA2020, predicting DNA-RNA-and protein-protein binding interface residues [4]. The same rationale held with respect to the prediction of effects upon molecular protein function (Table 1) [5].…”

Section: Table 1 Data Sets With Experimental Annotationsmentioning

confidence: 99%

“…Therefore, we included all known 6,699,150 SAVs from 60,706 people [5]. For all SAVs two prediction methods were applied: SNAP2 [15,16] predicted the effect of each SAV on molecular protein function, and ProNA2020 [4] predicted whether or not that SAV is in a ProNA-binding interface.…”

Section: Table 1 Data Sets With Experimental Annotationsmentioning

confidence: 99%

“…All results were based on the ExAC data from 60 k individuals [5]; SNAP2 [15,16] predicted effects on molecular protein function, and ProNA2020 [4] predicted residues at ProNA-binding interfaces (binding either other proteins, DNA, or RNA). (a demonstrates the degree to which SAVs (Single Amino acid Variants) are predicted more or less often than expected by chance (Methods) in ProNA-binding interfaces by the method ProNA2020 [4]. In particular, common SAVs (observed in > 5% of population) and rare SAVs (observed in < 1% of population) were significantly under-represented in ProNA-binding.…”

Section: Savs Binding Residues Under-representedmentioning

confidence: 99%

“…The differences between all pairwise curves were statistically significant (Additional File 1: Table S1). For instance, for very reliable effect predictions with SNAP2-scores ≥ 50 (dashed vertical lines), about 40% of all common SAVs were predicted to affect molecular function and to be in a residue predicted or observed (ProNA2020 [4] uses whatever is available, either a homology-based inference from experimental information or machine learning prediction) to be in an interface binding a large molecule (protein, DNA, or RNA) Table S3, note this defined the limit of the calculation using the software environment R [22]). The same trend held for each of the type of ProNA-binding, namely for protein, DNA, and RNA binding (Additional File 1: Table S3).…”

Section: Savs Binding Residues Under-representedmentioning

confidence: 99%

See 2 more Smart Citations

Protein–protein and protein-nucleic acid binding residues important for common and rare sequence variants in human

2020

Self Cite

View full text Add to dashboard Cite

Background Any two unrelated people differ by about 20,000 missense mutations (also referred to as SAVs: Single Amino acid Variants or missense SNV). Many SAVs have been predicted to strongly affect molecular protein function. Common SAVs (> 5% of population) were predicted to have, on average, more effect on molecular protein function than rare SAVs (< 1% of population). We hypothesized that the prevalence of effect in common over rare SAVs might partially be caused by common SAVs more often occurring at interfaces of proteins with other proteins, DNA, or RNA, thereby creating subgroup-specific phenotypes. We analyzed SAVs from 60,706 people through the lens of two prediction methods, one (SNAP2) predicting the effects of SAVs on molecular protein function, the other (ProNA2020) predicting residues in DNA-, RNA- and protein-binding interfaces. Results Three results stood out. Firstly, SAVs predicted to occur at binding interfaces were predicted to more likely affect molecular function than those predicted as not binding (p value < 2.2 × 10–16). Secondly, for SAVs predicted to occur at binding interfaces, common SAVs were predicted more strongly with effect on protein function than rare SAVs (p value < 2.2 × 10–16). Restriction to SAVs with experimental annotations confirmed all results, although the resulting subsets were too small to establish statistical significance for any result. Thirdly, the fraction of SAVs predicted at binding interfaces differed significantly between tissues, e.g. urinary bladder tissue was found abundant in SAVs predicted at protein-binding interfaces, and reproductive tissues (ovary, testis, vagina, seminal vesicle and endometrium) in SAVs predicted at DNA-binding interfaces. Conclusions Overall, the results suggested that residues at protein-, DNA-, and RNA-binding interfaces contributed toward predicting that common SAVs more likely affect molecular function than rare SAVs.

show abstract