PubHawk
Chem Sci March 2020 PMID 34122839

Predicting retrosynthetic pathways using transformer-based models and a hyper-graph exploration strategy.

Philippe Schwaller, Riccardo Petraglia, Valerio Zullo, Vishnu H Nair, Rico Andreas Haeuselmann, Riccardo Pisoni, Costas Bekas, Anna Iuliano, Teodoro Laino

Download social card

Mentions

36
Peak: 23 in Mar 2020
across 3 sources since Mar 2020

Extended reach

165,890
potential audience

Total likes

95

Video views

8,780

Mentions per month

Mar 2020 – Sep 2022

Where it's discussed

  • X / Twitter30 posts
  • Blogs4 articles
  • YouTube2 videos

AI synopsis

What people are saying

Many posts emphasize the innovative integration of transformer-based models for chemical synthesis planning, with phrases like 'automate the RoboRXN pipeline' and 'surpassed 3 million predictions,' indicating active interest and recognition of the publication's relevance. The tone varies from technical enthusiasm to curiosity about applications, which collectively signal a positive perception of the research's potential impact and importance.

AgreementModerate agreement

Most posts express support or enthusiasm for the advances described in the publication, highlighting connections to existing workflows and tools.

InterestHigh interest

Posts show high curiosity about practical applications and details of the models, reflecting active engagement with the technical content.

EngagementHigh engagement

Several discussions delve into the methodology, mentioning transformer models, hypergraph strategies, and accessibility of tools, indicating deep involvement.

ImpactModerate impact

While some posts underscore significance for the field, the overall perception is that the work has meaningful but not revolutionary impact at this stage.

Abstract in brief

  • The researchers developed an advanced computer model combining a Molecular Transformer with a hypergraph exploration strategy to automatically plan chemical synthesis routes without human help, improving how reactions are predicted step-by-step.
  • They introduced four new ways (metrics) to evaluate how well these models perform—covering aspects like how much of the possible reactions they can predict (coverage), the diversity of their predictions, the accuracy of the entire reaction process (roundtrip accuracy), and how similar the predictions are to known data (Jensen-Shannon divergence).
  • The model was tested with real examples, showing very good results, though some challenges remain—mainly related to the quality of the training data—while these evaluation metrics can help in fine-tuning future models for better performance.]
Read the full abstract

Top posts

Most liked on X and Bluesky
RXN for Chemistry
@ForRxn · X

The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 3, 2020 · 30 likes · 8 shares · 949 followers View post →
Philippe Schwaller (he/him)
@pschwllr · X

A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.co/qtAtK8EWB5)! Brilliant @IBMResearch and @Unipisa collaboration! How would you evaluate multi-step retrosynthetis models? #RSCOrg @teodorolaino @forRXN https://t.co/k8o407qqZf

Mar 4, 2020 · 27 likes · 13 shares · 5,281 followers View post →
Jannis Born
@JannisBorn · X

5/ Step 3: Select 250 best molecules & assess synthesizability with the multistep retrosynthesis model by @pschwllr/@ForRXN (https://t.co/P1AciYnK4H). 30% of the molecules were synthesizable, 50% of them in <= 2 steps. We then selected our final candidate: 3-Bromobenzylamine https://t.co/znPSD2YTti

Apr 15, 2021 · 7 likes · 1 share · 669 followers View post →
Philippe Schwaller (he/him)
@pschwllr · X

@kamerlinlab @Ben_Hoov @hen_str @teodorolaino @reymondgroup @IOPPublishing We investigated chemical reaction prediction (https://t.co/usDVueAuC4) and multi-step retrosynthesis planning (https://t.co/qtAtK8EWB5) in other studies. These tools can freely be accessed through https://t.co/7UkKVj7raT ->@forrxn. 4/4

Jul 16, 2020 · 5 likes · 1 share · 5,281 followers View post →
Philippe Schwaller (he/him)
@pschwllr · X

@andrewwhite01 @liubowen16 Multi-step synthesis planning using two transformers (retro and path scoring): https://t.co/dONTF8ylFw Our models are freely accessible (also without coding experience) through the IBM RXN for chemistry platform: https://t.co/7UkKVj7raT We recently surpassed 3M predictions!

Feb 24, 2021 · 2 likes · 5,281 followers View post →
Chemical Science
@ChemicalScience · X

RT @ForRxn: The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 4, 2020 · 8 shares · 99,649 followers View post →

All mentions

Teodoro Laino
@teodorolaino · X

RT @acvaucher: This was the missing piece to automate the RoboRXN pipeline: it now links the AI-suggested retrosynthetic route for a target…

May 6, 2021 · 1 share · 1,917 followers View post →
Nil Adell Mill
@fluck_5 · X

RT @JannisBorn: 5/ Step 3: Select 250 best molecules & assess synthesizability with the multistep retrosynthesis model by @pschwllr/@ForRX…

Apr 15, 2021 · 1 share · 426 followers View post →
Jannis Born
@JannisBorn · X

5/ Step 3: Select 250 best molecules & assess synthesizability with the multistep retrosynthesis model by @pschwllr/@ForRXN (https://t.co/P1AciYnK4H). 30% of the molecules were synthesizable, 50% of them in <= 2 steps. We then selected our final candidate: 3-Bromobenzylamine https://t.co/znPSD2YTti

Apr 15, 2021 · 7 likes · 1 share · 669 followers View post →
Philippe Schwaller (he/him)
@pschwllr · X

@andrewwhite01 @liubowen16 Multi-step synthesis planning using two transformers (retro and path scoring): https://t.co/dONTF8ylFw Our models are freely accessible (also without coding experience) through the IBM RXN for chemistry platform: https://t.co/7UkKVj7raT We recently surpassed 3M predictions!

Feb 24, 2021 · 2 likes · 5,281 followers View post →
RXN for Chemistry
@ForRxn · X

RT @pschwllr: @kamerlinlab @Ben_Hoov @hen_str @teodorolaino @reymondgroup @IOPPublishing We investigated chemical reaction prediction (http…

Jul 16, 2020 · 1 share · 949 followers View post →
Philippe Schwaller (he/him)
@pschwllr · X

@kamerlinlab @Ben_Hoov @hen_str @teodorolaino @reymondgroup @IOPPublishing We investigated chemical reaction prediction (https://t.co/usDVueAuC4) and multi-step retrosynthesis planning (https://t.co/qtAtK8EWB5) in other studies. These tools can freely be accessed through https://t.co/7UkKVj7raT ->@forrxn. 4/4

Jul 16, 2020 · 5 likes · 1 share · 5,281 followers View post →
aNyFuTuRe
@aNyFuTuRis · X

Predicting retrosynthetic pathways using transformer-based models and a hyper-graph exploration strategy - now published in Chemical Science https://t.co/z0exUujvu1

Apr 13, 2020 · 15 followers View post →
Ahmad Chandra
@chandra__21 · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 5, 2020 · 13 shares · 104 followers View post →
Amol Thakkar
@av_thakkar · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 586 followers View post →
Elif
@ozkirimli_elif · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 1,188 followers View post →
Chemical Science
@ChemicalScience · X

RT @ForRxn: The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 4, 2020 · 8 shares · 99,649 followers View post →
Matteo Manica
@drugilsberg · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 1,056 followers View post →
全合成たん
@chemiegg · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 2,473 followers View post →
おじさん
@snowborderjack · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 249 followers View post →
Mufei Li
@Mufei_Li · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 499 followers View post →
Thomas Struble
@thomasstruble · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 503 followers View post →
yamasaKit_
@yamasaKit_ · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 695 followers View post →
Teodoro Laino
@teodorolaino · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 1,917 followers View post →
RXN for Chemistry
@ForRxn · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 949 followers View post →
Michel VdP #StandWithUkraine #bettertogether
@michelvdp · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 2,061 followers View post →
Leonid Leiva
@turiciencia · X

RT @pschwllr: A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.c…

Mar 4, 2020 · 13 shares · 486 followers View post →
Philippe Schwaller (he/him)
@pschwllr · X

A last minute entry to #RSCPoster. Our Retrosynthetic Pathway Prediction Workflow - just out in @ChemicalScience (https://t.co/qtAtK8EWB5)! Brilliant @IBMResearch and @Unipisa collaboration! How would you evaluate multi-step retrosynthetis models? #RSCOrg @teodorolaino @forRXN https://t.co/k8o407qqZf

Mar 4, 2020 · 27 likes · 13 shares · 5,281 followers View post →
Amol Thakkar
@av_thakkar · X

RT @ForRxn: The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 3, 2020 · 8 shares · 586 followers View post →
Matteo Manica
@drugilsberg · X

RT @ForRxn: The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 3, 2020 · 8 shares · 1,056 followers View post →
Agenparl
@Agenparl · X

Predicting #Retrosynthetic Pathways using Transformer-Based Models and a Hyper-Graph Exploration Strategy - https://t.co/OOV6gwGiwd #Agenparl #Class #Model #Readability #Single #AgenparlEnglish #Educazione #ScienzaETecnologia #SocialNetwork https://t.co/LQF2uyTZr7

Mar 3, 2020 · 6,063 followers View post →
RSC_CICAG
@RSC_CICAG · X

RT @ForRxn: The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 3, 2020 · 8 shares · 1,618 followers View post →
Chris Swain
@macinchem · X

RT @ForRxn: The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 3, 2020 · 8 shares · 1,537 followers View post →
Daniel Probst @skepteis@skepteis.social
@skepteis · X

RT @ForRxn: The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 3, 2020 · 8 shares · 10,649 followers View post →
Teodoro Laino
@teodorolaino · X

RT @ForRxn: The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 3, 2020 · 8 shares · 1,917 followers View post →
RXN for Chemistry
@ForRxn · X

The details on the retrosynthetic engine driving the design of chemical synthesis ⚗️

Mar 3, 2020 · 30 likes · 8 shares · 949 followers View post →